JNS-13862; No of Pages 5 Journal of the Neurological Sciences xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Multiple sclerosis and breast cancer P.W. O'Malley a,b,1, Z.D. Mulla c,d,1, O. Nesic e,⁎ a
Texas Tech University Health Sciences Center El Paso, Paul L. Foster School of Medicine, United States University of Texas School of Public Health at Houston, United States c Department of Obstetrics and Gynecology, Texas Tech University Health Sciences Center El Paso, Paul L. Foster School of Medicine, TX, United States d Department of Public Health, Graduate School of Biomedical Sciences, Texas Tech University Health Sciences Center, Lubbock, TX, United States e Department of Medical Education, Texas Tech University Health Sciences Center El Paso: Paul L. Foster School of Medicine, United States b
a r t i c l e
i n f o
Article history: Received 5 May 2015 Received in revised form 30 May 2015 Accepted 13 June 2015 Available online xxxx Keywords: Multiple sclerosis Breast cancer Case–control study Odds ratios Texas hospital data Age Race Health insurance Diabetes Open wound
a b s t r a c t Multiple sclerosis (MS) and breast cancer (BC) share common features; most notably, both are more frequent in women than in men. In addition to the involvement of sex hormones, a number of genetic and pharmacological studies support a possible relationship between these two diseases. However, there are no conclusive epidemiological findings related to MS and BC worldwide, and there are no recent data for the US population. We conducted a case–control study using a hospital inpatient discharge dataset (21,536 cases and two control series totaling 59,581 controls) from the Texas Health Care Information Collection. We assessed occurrence of MS in BC cases and in two control series: diabetes mellitus type II, and open wounds. After controlling for age, race-ethnicity, and health insurance status, a statistically-significant protective association was detected: BC cases were 45% less likely than diabetic controls to have MS (OR = 0.55, 95% CI = 0.37–0.81), and 63% less likely than open wound controls to have MS (OR = 0.37, 95% CI = 0.21–0.66). Our study presented here is the only current assessment of the association between MS and BC in the USA and suggests a protective effect of MS on BC in the hospitalized population. Published by Elsevier B.V.
1. Introduction Multiple sclerosis (MS) is a demyelinating disease affecting 400,000 people in the United States [1]. The prevalence of MS is higher in women (3.2:1), suggesting a potential role of sex hormones and/or sex chromosomes in the pathogenesis or progression of the disease [2,3]. Interestingly, both estrogens and androgens appear to have a protective role in multiple sclerosis [4,5]. Although the role of sex hormones in MS remains incompletely understood, it raises questions regarding comorbidities with other diseases involving sex hormones, such as breast cancer (BC), given that higher estrogen levels are implicated in breast cancer for both pre-and post-menopausal women [6,7], and that BC treatments often involve drugs that target estrogen receptors [8] or estrogen synthesis [9]. However, despite compelling endocrinological, pharmacological, and genetic [10] evidences that strongly link MS and BC, we currently lack a conclusive epidemiological answer to whether MS increases or decreases risk for BC. Few population-based studies have been ⁎ Corresponding author at: Texas Tech University Health Science Center Paul L. Foster School of Medicine, 5001 El Paso Dr. El Paso, TX 79905, United States. E-mail address: [email protected] (O. Nesic). 1 Authors contributed the same.
conducted in the U.S. investigating the comorbidity of MS and BC; one in Minnesota (1975 to 1984) reports a prevalence of BC in MS patients of 2.01% [11]. The other studies performed in the US report either decreased prevalence of BC in hospitalized MS patients [25] or reduced comorbidity of BC in self-reports of MS patients evaluated through the NARCOMS registry (NARCOMS: North American Research Committee on Multiple Sclerosis) [12]. However, a systematic review of published findings assessing risks of BC in MS patients evaluated in different countries/continents, found much variation. The dearth of conclusive epidemiological findings related to MS and BC worldwide and the absence of current data for the US population compelled us to investigate the relationship between MS and BC using data collected by the Texas Health Care Information Collection, Center for Health Statistics. 2. Materials and methods 2.1. Source population and inclusion criteria A case–control study was conducted using the Texas Public Use Data File, a hospital inpatient discharge dataset from the Texas Health Care Information Collection (THCIC), Texas Department of State Health Services (Austin, Texas). THCIC data are from all state licensed hospitals
http://dx.doi.org/10.1016/j.jns.2015.06.033 0022-510X/Published by Elsevier B.V.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
2
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
except those that are exempt from reporting to the THCIC. Exempt hospitals include those located in a county with a population less than 35,000, or those located in a county with a population more than 35,000 and with fewer than 100 licensed hospital beds and not located in an area that is delineated as an urbanized area by the United States Bureau of the Census (Section 108.0025). Exempt hospitals also include hospitals that do not seek insurance payment or government reimbursement (Section 108.009). According to our university's Institutional Review Board Policies and Procedures Manual 1.4.1.2: “Research using unidentifiable publicly or commercially available databases, human cell lines, or material from human cadavers is not considered to meet the definition of a human subject, and, as such, does not require IRB review or approval.” The dataset that was available in our institution contained clinical and demographic information for patients who were discharged in calendar years 2004 through 2007. Because of the relatively low prevalence of MS we analyzed data collected over four consecutive years. The principal discharge field and 24 secondary discharge diagnosis fields were examined in our study. The discharge variables had been coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). The records of women who were less than 20 years of age or who had their age classified in one of several broad age categories due to HIV infection and/or drug and/or ethanol use were excluded. 2.2. Definition of cases and controls BC was the outcome of interest. Cases were women whose principal discharge diagnosis field contained an ICD-9-CM code beginning with 174 (malignant neoplasm of the female breast). Although the frequency of BC among different races in our data base (biased towards white nonHispanics) was different than in the general population, the age distribution (see Table 1) closely matched the bell-shaped age distribution of BC cases in the general population (National Cancer Institute: Breast cancer). To strengthen our analyses we used two control series: (1) “controls series 1” were women whose principal discharge diagnosis code began with 250 and ended in 0 or 2 (code 250.XX identifies patients with diabetes mellitus while the fifth digit of 0 or 2 identifies patients with type II diabetes mellitus) and who did not have an ICD-9-CM code beginning with 174 in any of her secondary discharge diagnosis fields, and (2) “controls series 2” was composed of women whose principal
discharge diagnosis code began with 870 through 887 or 890 through 897 (open wound) and who did not have an ICD-9-CM code beginning with 174 in any of their secondary discharge diagnosis fields. We have chosen two control conditions that have not been associated with MS. Although the prevalence of Type I diabetes is 3-fold greater in the MS population [14], no association between diabetes type II and MS has been found [15]. For our second control series we have chosen patients hospitalized with the primary diagnosis of open wounds, as there is no evidence that there is any association between MS and open wounds. The use of two control series (a chronic condition, that is, Type II diabetes, and an acute condition, open wounds) with varying hospital admission probabilities most likely reduces the risk that our results were wholly due to Berkson's bias [16]. 2.3. Definition of MS MS was the main exposure variable. MS was defined as the presence of an ICD-9-CM code beginning with 340 in any of the secondary discharge diagnosis fields. Patients who did not have a code beginning with 340 in any of their secondary discharge diagnosis fields were considered to be free of MS. 2.4. Data analysis Data were analyzed using SAS 9.3 software (SAS Institute, Inc., Cary, North Carolina). Initial analyses involved the creation of contingency tables with a significance level of 0.05. Unadjusted and adjusted odds ratios (OR), 95% confidence intervals (CI), and P values were calculated from unconditional logistic regression models. ORs for the association between MS and BC were adjusted for the patient's age (seven age groups modeled using six indicator variables), the patient's raceethnicity (four groups), and the patient's health insurance information (3 groups). The original health insurance variable found in our dataset has over 20 possible response values. We collapsed these categories and created a new health insurance variable with the following three groups: (1) self-pay or indigent (combined together, see Anderson, 2007 [17]), (2) Medicaid, and (3) cases that are not in either group (1) or (2). Parity is another possible confounding factor in any analysis in which the outcome is BC and MS since nulliparous women have a higher risk of developing BC [18] and nulliparity is more frequent among MS patients [19]. We sought to control for parity indirectly by adjusting
Table 1 Characteristics of the study sample. Female breast cancer (BC) cases were compared with female type II diabetic controls (DC) and female open wound (OW) controls. Controls did not have a secondary discharge diagnosis of BC. The patients were discharged throughout Texas between 2004 and 2007 and found in the Public Use Data File. Variable
BC cases
DC
P
OW controls
BC cases vs. DC
Age (years) 20–29 30–39 40–49 50–59 60–69 70–79 ≥80 Race-ethnicity Black non-Hispanic White Hispanic White non-Hispanic Other (Asian, Native American, etc.) Health insurance Self-pay, indigent Medicaid Other Has multiple sclerosis
N = 21,536
N = 54,141
Number (%)
Number (%)
BC cases vs. OW Controls N = 5440 Number (%)
b0.0001 156 (0.72) 1348 (6.3) 4394 (20.4) 5498 (25.5) 4509 (20.9) 3490 (16.2) 2141 (9.9)
1443 (2.7) 3729 (6.9) 7959 (14.7) 12,212 (22.6) 11,198 (20.7) 10,369 (19.2) 7231 (13.4)
2812 (13.1) 1355 (6.3) 14,015 (65.1) 3354 (15.6)
12,955 (23.9) 7550 (14.0) 19,497 (36.0) 14,139 (26.1)
1574 (7.3) 1309 (6.1) 18,653 (86.6) 37 (0.17)
6799 (12.6) 6582 (12.2) 40,760 (75.3) 103 (0.19)
P
b0.0001 811 (14.9) 721 (13.3) 888 (16.3) 783 (14.4) 540 (9.9) 634 (11.7) 1063 (19.5)
b0.0001
b0.0001 687 (12.6) 511 (9.4) 3285 (60.4) 957 (17.6)
b0.0001
0.59
b0.0001 957 (17.6) 395 (7.3) 4088 (75.2) 17 (0.31)
0.04
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
for the patient's age since further analysis of the study by Mulla et al. demonstrated that nulliparity at the time of presentation to a labor and delivery unit among women delivering at a single hospital declined steadily with increasing age of the woman, and hence a woman's age can be a proxy measure for her parity [20]. We also analyzed the frequency distribution of the principal procedure field among BC cases. If an adjusted MS OR differed by 10 or more percent from the crude MS OR, then confounding was considered to be present [21].
3
63% less likely than OW controls to have MS (adjusted OR = 0.37, P = 0.001). We attempted to conduct a second case–control study among men (ICD9 code 175, malignant neoplasm of the male breast; n = 136 with eligible values for patient age, race-ethnicity, and health insurance), but none had a secondary discharge diagnosis code for MS so further analyses were not completed. 4. Discussion
3. Results The flowchart in Fig. 1 presents frequency distribution of the principal procedure code for our BC cases; i.e. the number of cases and controls taken into our analyses. We identified 21,600 female cases with the principal discharge diagnosis of breast cancer (BC). However, 124 patients were removed from the study because of invalid data for age, race/ethnicity or health insurance information. As a result, 21,536 cases were taken into the further analysis. Similar evaluation of data for two control series yielded 54,141 patients with diabetes type II and 5440 patients with open wounds. Characteristics of the study sample are reported in Table 1. A statistically-significant difference in the age distribution was found between BC cases and diabetic controls (DC) (P b 0.0001), and BC cases and open wound (OW) controls (P b 0.0001). The prevalence of MS was 0.17% in BC cases and 0.19% (P = 0.59) and 0.31% (P = 0.04) in diabetic and open wounds controls, respectively. The frequency distribution of the principal procedure that was performed during the hospital stay of the BC patients was also examined. This analysis showed that 77% of the BC cases had undergone surgical procedures for the management of breast cancer. For example, 9741 BC patients had unilateral extended simple mastectomy (ICD-9 CM code 85.43), while 2897 BC patients had unilateral simple mastectomy (ICD-9 CM code 85.41), procedures that are usually performed early after BC diagnosis, suggesting that most of the BC cases that were included in our analysis were likely newly diagnosed BC cases. Unadjusted (crude) and adjusted ORs for the relationship between MS and the outcome of BC are shown in Table 2. Both adjusted ORs differ by ≥10% from their respective unadjusted ORs. After controlling for age, race-ethnicity, and health insurance status, a statistically-significant protective association was detected: BC cases were 45% less likely than DC controls to have MS (adjusted OR = 0.55, P = 0.003), and
Our novel analysis of a population-based hospital dataset found an inverse association between MS and the outcome of BC, suggesting a decreased risk for BC in MS patients, given that in the majority of cases MS precedes BC: the average age of MS diagnosis is 36 years [22] and for BC diagnosis the average age is 61 [23]. Breast cancer diagnosed in women 35 years of age or less accounts for b 2% of all breast cancer cases [24] further supporting our assumption that MS precedes BC in majority of cases. Two other studies also found a decreased prevalence and incidence of BC in MS patients, in US and France, respectively [25,26]. However, conflicting findings abound in the current literature regarding the comorbidity of BC and MS in other countries. For example, some studies found an increased incidence of BC in MS patients in Taiwan or Norway, while other studies found no significant association between MS and BC in Finland, Israel, or Canada (for a detailed review see Marrie et al., 2015 [13]). As discussed in Marrie et al. [13], the literature characterizing the risk of all cancer (including BC) in MS is typified by incongruent study designs and a lack of consistent adjustment for various confounders, and thus can explain the observed variability among different reports. It is also possible that the presence or absence of association between MS and BC in various populations that were analyzed at different times reflects the variability in pharmacological treatments of analyzed MS patients. There are currently at least seven different diseasemodifying treatments for MS (reviewed in Files et al., 2015 [27]), most of which become available only recently. Our study adjusted for age, race and health insurance, and used a comprehensive statewide hospital discharge database obtained from the THCIC, rather than analyzing data from a single institution or a limited geographical area. In 2007, quarterly figures indicated that on average 80% of the state licensed hospitals in Texas were required to report their data to the THCIC and of this group 99% did report (Personal communication, THCIC staff, Austin, Texas). Furthermore, Texas is a state
Fig. 1. Numbers of individuals in our study sample. *Cases were deemed invalid if the values for age, and/or race/ethnicity and/or health insurance were ineligible, missing, or nonsensical.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
4
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
Table 2 Unadjusted and adjusted odds ratios (OR), 95% confidence intervals (CI), and P values for the association between multiple sclerosis (MS), present vs. absent, and the outcome of breast cancer (BC) in women discharged throughout Texas. BC cases (N = 21,536) were compared to female type II diabetic controls (DC) (N = 54,141) and female open wound (OW) controls (N = 5440). controls did not have a secondary discharge diagnosis of BC. Type of OR
DC MS OR (95% CI)
Unadjusted Adjusted for the patient's age, race-ethnicity, and health insurance
OW controls P
0.91 0.60 (0.62–1.32) 0.55 0.003 (0.37–0.81)
MS OR (95% CI)
P
0.55 0.04 (0.31–0.97) 0.37 0.001 (0.21–0.66)
that is large both in terms of population (estimated at 22,778,123 in 2005; United States Census Bureau) and size (land area in 2010 was approximately 261,232 mile2; Texas QuickFacts; U.S. Census bureau: 2015) indicating that our study involved a large geographical area and a large number of patients: 21,536 cases and 59,581 controls. Although diagnoses found in hospital discharge datasets may suffer from miscoding, a review of the validity of the ICD coding of neurological conditions in administrative health data by St. Germaine Smith et al. (2012) found that the coding of MS was accurate (sensitivity ranging from 85 to 92.4%, and a positive predictive value of 74.5 to 92.7%) [28]. Another possible, but unlikely limitation of our study stems from the fact that the Texas Inpatient Public Use Data File (PUDF) does not include a unique patient identifier, and hence there is a chance that multiple records were included in our sample for one or more patients. It is possible that patients were transferred between two hospitals both of whom report to the PUDF. However, our analysis showed that only 1.6% of our BC cases, 4.5% of our diabetic controls, and 6.8% of our OW controls had an unknown value for their source of admission variable or had been identified as transfers from other hospitals. Furthermore, it is possible that the initial admission was at a smaller facility that is not required to report their patient discharge data to the PUDF and hence only one record for the patient was included in our sample. Therefore, this limitation of de-identified hospital inpatient discharge dataset unlikely affected the result of our study. It is also possible that an individual was discharged to their home and was re-admitted to the same facility due to a complication; however, in this situation it is likely that the complication, e.g., an infection, rather than BC, would be noted in the principal discharge diagnosis field. Potential interpretations of the inverse association between BC and MS obtained in our study include the possibility that MS treatments of analyzed patients decreased the risk for developing BC, conceivable common biological mechanisms, and/or that the biased selection of hospitalized cases and controls influenced the outcome in our study. Although we could not analyze MS treatments (data were not available), it is possible that MS drugs influence the development of BC, as suggested in some studies. For example, an Israeli study reports that interferon beta (IFNβ), which has been approved for the treatment of MS since 1993 (and thus it was likely used by MS patients in our study) lowers the risk of BC in female patients [29]. Furthermore, recent studies also demonstrate that IFNβ has an import role in the BC pathology [30], and IFNβ is now considered as a possible intervention for BC [31]. Despite evident clinical relevance, epidemiological studies addressing the effect of MS treatments on BC are scarce and inconclusive [32]. In kind, immunomodulatory treatments that are potentially beneficial for both MS and BC may imply common biological mechanisms underlying both diseases. Nevertheless, similar immune reactions [33] and the interplay of sex-specific factors or the genetic background [34] linking MS and BC cannot fully explain a potentially protective effect of MS on BC. We attempted to shed light on the possible contribution of the sex-specific factors to the risk of BC in MS by also analyzing male population in our data base. However, given that BC is 100 times less common in men [35], we identified only 136 men with BC, and
since MS is more than twice as common in women than in men [36], we expectedly did not identify any men with both BC and MS. Although the comparison between BC risks in men and women with MS has not yet been reported, Marrie et al. have shown that the percent of selfreported BC in men with MS (3%) is higher than in women with MS (2.3%), indicating a possibility that sex-specific factors contribute to the risk of BC in MS patients [12]. Any hospital-based case–control study in which both the exposure and the outcome are diseases may suffer from Berkson's bias, [16,37]. We attempted to minimize the possibility that our results were wholly due to Berkson's bias by using two control conditions, one acute and one chronic, which may have different hospital admission probabilities. The use of multiple control series can be viewed as a form of sensitivity analysis. Our estimation of the strength of the association between MS and BC was largely insensitive to the choice of the control disease/condition: both of the adjusted exposure ORs (Table 2) indicated the presence of a strong, protective relationship. Furthermore, because the use of prevalent cases may lead to biased ORs due to the effect of selective survival, we confirmed that the majority of our BC cases were newly diagnosed (incident), thereby reducing the possibility of selective survival bias. Even though cases in our study closely matched the age distribution of BC in general population, the race/ethnicity distribution did not, with 2/3 of analyzed women being white non-Hispanics, thus reflecting unbalanced admission probabilities and/or unequal BC incidence (at least for different races/ethnicities). Therefore, it is possible that marked underrepresentation of black women in our study affected the overall association between MS and BC, as black women have higher incidence of BC than white women [35]. This implication raises an interesting possibility that the risk for BC in MS patients depends on the race/ethnicity, which remains to be investigated further. Our study presented here is the only recent assessment of the association between MS and BC in a relatively large area of the US. It is worth noting that only two reports [11,25] that analyzed MS and BC association in the US preceded ours by a few decades; i.e. analyzed data were collected from 1975 to 1984, or in 1989, respectively. In addition, those two studies investigated more restricted populations, i.e. in the Olmsted County (Minnesota), or only Medicare patients older than 65. Our results suggest a protective effect of MS in the hospital-based population of BC patients, in agreement with earlier reports [25,26]. Given the clinical implication of such association, both more comprehensive epidemiological investigations of the USA population, and studies analyzing biological underpinnings of potentially protective effects of MS are sorely needed.
Conflict of interest None exists.
Acknowledgments This work was in part supported by the 2014 Paul L. Foster School of Medicine Individual Investigator-Initiated Seed Grant Program (P.I. ON). The authors acknowledge the use of the Texas Hospital Inpatient Discharge Public Use Data File, quarters 1 through 4, for the years 2004 through 2007, Texas Department of State Health Services, Center for Health Statistics-Texas Health Care Information Collection, Austin, Texas, USA.
Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jns.2015.06.033.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
References [1] M.J. Tullman, Overview of the epidemiology, diagnosis, and disease progression associated with multiple sclerosis, Am. J. Manag. Care 19 (2 Suppl.) (2013) S15–S20. [2] L. Moroni, I. Bianchi, A. Lleo, Geoepidemiology, gender and autoimmune disease, Autoimmun. Rev. 11 (6–7) (2012) A386–A392. [3] M. Kipp, S. Amor, R. Krauth, C. Beyer, Multiple sclerosis: neuroprotective alliance of estrogen-progesterone and gender, Front. Neuroendocrinol. 33 (1) (2012) 1–16. [4] R.D. Spence, R.R. Voskuhl, Neuroprotective effects of estrogens and androgens in CNS inflammation and neurodegeneration, Front. Neuroendocrinol. 33 (1) (2012) 105–115. [5] B.F. Bebo Jr., E. Zelinka-Vincent, G. Adamus, D. Amundson, A.A. Vandenbark, H. Offner, Gonadal hormones influence the immune response to PLP 139–151 and the clinical course of relapsing experimental autoimmune encephalomyelitis, J. Neuroimmunol. 84 (2) (1998) 122–130. [6] Group EHBCC, Body mass index, serum sex hormones, and breast cancer risk in postmenopausal women, J. Natl. Cancer Inst. 95 (16) (2003) 1218–1226. [7] R. Kaaks, K. Tikk, D. Sookthai, H. Schock, T. Johnson, A. Tjønneland, et al., Premenopausal serum sex hormone levels in relation to breast cancer risk, overall and by hormone receptor status—results from the EPIC cohort, Int. J. Cancer 134 (8) (2014) 1947–1957. [8] N. Heldring, A. Pike, S. Andersson, J. Matthews, G. Cheng, J. Hartman, et al., Estrogen receptors: how do they signal and what are their targets, Physiol. Rev. 87 (3) (2007) 905–931. [9] S. Gobbi, A. Rampa, F. Belluti, A. Bisi, Nonsteroidal aromatase inhibitors for the treatment of breast cancer: an update, Anti Cancer Agents Med. Chem. 14 (1) (2014) 54–65. [10] C. Holzmann, I. Bauer, P. Meyer, Co-occurrence of multiple sclerosis and cancer in a BRCA1 positive family, Eur. J. Med. Genet. 56 (10) (2013) 577–579. [11] D.R. Wynn, M. Rodriguez, W.M. O'Fallon, L.T. Kurland, A reappraisal of the epidemiology of multiple sclerosis in Olmsted County, Minnesota, Neurology 40 (5) (1990) 780–786. [12] R. Marrie, R. Horwitz, G. Cutter, T. Tyry, D. Campagnolo, T. Vollmer, Comorbidity, socioeconomic status and multiple sclerosis, Mult. Scler. 14 (8) (2008) 1091–1098. [13] R.A. Marrie, N. Reider, J. Cohen, O. Stuve, M. Trojano, P.S. Sorensen, et al., A systematic review of the incidence and prevalence of cancer in multiple sclerosis, Mult. Scler. 21 (3) (2015) 294–304. [14] P. Tettey, S. Simpson Jr., B.V. Taylor, I.A.F. van der Mei, The co-occurrence of multiple sclerosis and type 1 diabetes: shared aetiologic features and clinical implication for MS aetiology, J. Neurol. Sci. 348 (1–2) (2015) 126–131. [15] G.S. Watson, S. Craft, Insulin resistance, inflammation, and cognition in Alzheimer's Disease: lessons for multiple sclerosis, J. Neurol. Sci. 245 (1–2) (2006) 21–33. [16] J. Schwartzbaum, A. Ahlbom, M. Feychting, Berkson's bias reviewed, Eur. J. Epidemiol. 18 (12) (2003) 1109–1112. [17] G.F. Anderson, From ‘Soak The Rich’ To ‘Soak The Poor’: recent trends in hospital pricing, Health Aff. 26 (3) (2007) 780–789. [18] J. Russo, G.A. Balogh, I.H. Russo, Full-term pregnancy induces a specific genomic signature in the human breast, Cancer Epidemiol. Biomark. Prev. 17 (1) (2008) 51–66. [19] P. Cavalla, V. Rovei, S. Masera, M. Vercellino, M. Massobrio, R. Mutani, et al., Fertility in patients with multiple sclerosis: current knowledge and future perspectives, Neurol. Sci. 27 (4) (2006) 231–239.
5
[20] Z.D. Mulla, T. Carrillo, R. Kalamegham, L.L. Hernandez, E. Portugal, B.S. Nuwayhid, Is maternal colonization with group B streptococci a risk factor for preeclampsia? J. Reprod. Med. 60 (3–4) (2015) 117–126. [21] S. Greenland, Modeling and variable selection in epidemiologic analysis, Am. J. Public Health 79 (3) (1989) 340–349. [22] M.L. Sumelahti, E. Pukkala, M. Hakama, Cancer incidence in multiple sclerosis: a 35year follow-up, Neuroepidemiology 23 (5) (2004) 224–227. [23] N. Howlander, A. Noone, M. Krapcho, J. Garshell, D. Miller, S. Altekruse, SEER Cancer Statistics Review, 1975–2011, National Cancer Institute, Bethesda, MD, 2014. [24] S. Bayraktar, L. Amendola, A.M. Gutierrez-Barrera, S.S. Hashmi, C. Amos, M. Gambello, Clinicopathologic characteristics of breast cancer in BRCA-carriers and non-carriers in women 35 years of age or less, The Breast 23 (6) (2014) 770–774. [25] S.T. Fleming, R.L. Blake Jr., Patterns of comorbidity in elderly patients with multiple sclerosis, J. Clin. Epidemiol. 47 (10) (1994) 1127–1132. [26] C. Lebrun, M. Debouverie, P. Vermersch, P. Clavelou, L. Rumbach, J. de Seze, et al., Cancer risk and impact of disease-modifying treatments in patients with multiple sclerosis, Mult. Scler. 14 (3) (2008) 399–405. [27] D.K. Files, T. Jausurawong, R. Katrajian, R. Danoff, Multiple sclerosis, Prim. Care 42 (2) (2015) 159–175. [28] C. St. Germaine-Smith, A. Metcalfe, T. Pringsheim, J.I. Roberts, C.A. Beck, B.R. Hemmelgarn, et al., Recommendations for optimal ICD codes to study neurologic conditions: a systematic review, Neurology 79 (10) (2012) 1049–1055. [29] A. Achiron, Y. Barak, M. Gail, M. Mandel, D. Pee, R. Ayyagari, et al., Cancer incidence in multiple sclerosis and effects of immunomodulatory treatments, Breast Cancer Res. Treat. 89 (3) (2005) 265–270. [30] H. Xu, J. Xian, E. Vire, S. McKinney, V. Wei, J. Wong, et al., Up-regulation of the interferon-related genes in BRCA2 knockout epithelial cells, J. Pathol. 234 (3) (2014) 386–397. [31] A. Carpi, A. Nicolini, A. Antonelli, P. Ferrari, G. Rossi, Cytokines in the management of high risk or advanced breast cancer: an update and expectation, Curr. Cancer Drug Targets 9 (8) (2009) 888–903. [32] E. Kingwell, C. Evans, F. Zhu, J. Oger, S. Hashimoto, H. Tremlett, Assessment of cancer risk with β-interferon treatment for multiple sclerosis, J Neurol Neurosurg Psychiatry 85 (10) (2014) 1092–1102. [33] M. Anagnostouli, G. Anagnostoulis, S. Katsavos, M. Panagiotou, E. Kararizou, P. Davaki, HLA-DRB1*15:01 and Epstein–Barr virus in a multiple sclerosis patient with psoriasis, nasopharyngeal and breast cancers. Lessons for possible hidden links for autoimmunity and cancer, J. Neurol. Sci. 339 (1–2) (2014) 26–31. [34] J.J. Chen, F. Costello, R.H. Kardon, Sex disparities in neuro-ophthalmologic disorders, Curr. Eye Res. 40 (2) (2015) 247–265. [35] B.A. Kohler, R.L. Sherman, N. Howlader, A. Jemal, A.B. Ryerson, K.A. Henry, et al., Annual report to the nation on the status of cancer, 1975–2011, featuring incidence of breast cancer subtypes by race/ethnicity, poverty, and state, J. Natl. Cancer Inst. 107 (6) (2015). [36] A. Alonso, M.A. Hernan, Temporal trends in the incidence of multiple sclerosis: a systematic review, Neurology 71 (2) (2008) 129–135. [37] J. Berkson, Limitations of the application of fourfold table analysis to hospital data, Biometrics 2 (3) (1946) 47–53.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
Contents lists available at ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Multiple sclerosis and breast cancer P.W. O'Malley a,b,1, Z.D. Mulla c,d,1, O. Nesic e,⁎ a
Texas Tech University Health Sciences Center El Paso, Paul L. Foster School of Medicine, United States University of Texas School of Public Health at Houston, United States c Department of Obstetrics and Gynecology, Texas Tech University Health Sciences Center El Paso, Paul L. Foster School of Medicine, TX, United States d Department of Public Health, Graduate School of Biomedical Sciences, Texas Tech University Health Sciences Center, Lubbock, TX, United States e Department of Medical Education, Texas Tech University Health Sciences Center El Paso: Paul L. Foster School of Medicine, United States b
a r t i c l e
i n f o
Article history: Received 5 May 2015 Received in revised form 30 May 2015 Accepted 13 June 2015 Available online xxxx Keywords: Multiple sclerosis Breast cancer Case–control study Odds ratios Texas hospital data Age Race Health insurance Diabetes Open wound
a b s t r a c t Multiple sclerosis (MS) and breast cancer (BC) share common features; most notably, both are more frequent in women than in men. In addition to the involvement of sex hormones, a number of genetic and pharmacological studies support a possible relationship between these two diseases. However, there are no conclusive epidemiological findings related to MS and BC worldwide, and there are no recent data for the US population. We conducted a case–control study using a hospital inpatient discharge dataset (21,536 cases and two control series totaling 59,581 controls) from the Texas Health Care Information Collection. We assessed occurrence of MS in BC cases and in two control series: diabetes mellitus type II, and open wounds. After controlling for age, race-ethnicity, and health insurance status, a statistically-significant protective association was detected: BC cases were 45% less likely than diabetic controls to have MS (OR = 0.55, 95% CI = 0.37–0.81), and 63% less likely than open wound controls to have MS (OR = 0.37, 95% CI = 0.21–0.66). Our study presented here is the only current assessment of the association between MS and BC in the USA and suggests a protective effect of MS on BC in the hospitalized population. Published by Elsevier B.V.
1. Introduction Multiple sclerosis (MS) is a demyelinating disease affecting 400,000 people in the United States [1]. The prevalence of MS is higher in women (3.2:1), suggesting a potential role of sex hormones and/or sex chromosomes in the pathogenesis or progression of the disease [2,3]. Interestingly, both estrogens and androgens appear to have a protective role in multiple sclerosis [4,5]. Although the role of sex hormones in MS remains incompletely understood, it raises questions regarding comorbidities with other diseases involving sex hormones, such as breast cancer (BC), given that higher estrogen levels are implicated in breast cancer for both pre-and post-menopausal women [6,7], and that BC treatments often involve drugs that target estrogen receptors [8] or estrogen synthesis [9]. However, despite compelling endocrinological, pharmacological, and genetic [10] evidences that strongly link MS and BC, we currently lack a conclusive epidemiological answer to whether MS increases or decreases risk for BC. Few population-based studies have been ⁎ Corresponding author at: Texas Tech University Health Science Center Paul L. Foster School of Medicine, 5001 El Paso Dr. El Paso, TX 79905, United States. E-mail address: [email protected] (O. Nesic). 1 Authors contributed the same.
conducted in the U.S. investigating the comorbidity of MS and BC; one in Minnesota (1975 to 1984) reports a prevalence of BC in MS patients of 2.01% [11]. The other studies performed in the US report either decreased prevalence of BC in hospitalized MS patients [25] or reduced comorbidity of BC in self-reports of MS patients evaluated through the NARCOMS registry (NARCOMS: North American Research Committee on Multiple Sclerosis) [12]. However, a systematic review of published findings assessing risks of BC in MS patients evaluated in different countries/continents, found much variation. The dearth of conclusive epidemiological findings related to MS and BC worldwide and the absence of current data for the US population compelled us to investigate the relationship between MS and BC using data collected by the Texas Health Care Information Collection, Center for Health Statistics. 2. Materials and methods 2.1. Source population and inclusion criteria A case–control study was conducted using the Texas Public Use Data File, a hospital inpatient discharge dataset from the Texas Health Care Information Collection (THCIC), Texas Department of State Health Services (Austin, Texas). THCIC data are from all state licensed hospitals
http://dx.doi.org/10.1016/j.jns.2015.06.033 0022-510X/Published by Elsevier B.V.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
2
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
except those that are exempt from reporting to the THCIC. Exempt hospitals include those located in a county with a population less than 35,000, or those located in a county with a population more than 35,000 and with fewer than 100 licensed hospital beds and not located in an area that is delineated as an urbanized area by the United States Bureau of the Census (Section 108.0025). Exempt hospitals also include hospitals that do not seek insurance payment or government reimbursement (Section 108.009). According to our university's Institutional Review Board Policies and Procedures Manual 1.4.1.2: “Research using unidentifiable publicly or commercially available databases, human cell lines, or material from human cadavers is not considered to meet the definition of a human subject, and, as such, does not require IRB review or approval.” The dataset that was available in our institution contained clinical and demographic information for patients who were discharged in calendar years 2004 through 2007. Because of the relatively low prevalence of MS we analyzed data collected over four consecutive years. The principal discharge field and 24 secondary discharge diagnosis fields were examined in our study. The discharge variables had been coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). The records of women who were less than 20 years of age or who had their age classified in one of several broad age categories due to HIV infection and/or drug and/or ethanol use were excluded. 2.2. Definition of cases and controls BC was the outcome of interest. Cases were women whose principal discharge diagnosis field contained an ICD-9-CM code beginning with 174 (malignant neoplasm of the female breast). Although the frequency of BC among different races in our data base (biased towards white nonHispanics) was different than in the general population, the age distribution (see Table 1) closely matched the bell-shaped age distribution of BC cases in the general population (National Cancer Institute: Breast cancer). To strengthen our analyses we used two control series: (1) “controls series 1” were women whose principal discharge diagnosis code began with 250 and ended in 0 or 2 (code 250.XX identifies patients with diabetes mellitus while the fifth digit of 0 or 2 identifies patients with type II diabetes mellitus) and who did not have an ICD-9-CM code beginning with 174 in any of her secondary discharge diagnosis fields, and (2) “controls series 2” was composed of women whose principal
discharge diagnosis code began with 870 through 887 or 890 through 897 (open wound) and who did not have an ICD-9-CM code beginning with 174 in any of their secondary discharge diagnosis fields. We have chosen two control conditions that have not been associated with MS. Although the prevalence of Type I diabetes is 3-fold greater in the MS population [14], no association between diabetes type II and MS has been found [15]. For our second control series we have chosen patients hospitalized with the primary diagnosis of open wounds, as there is no evidence that there is any association between MS and open wounds. The use of two control series (a chronic condition, that is, Type II diabetes, and an acute condition, open wounds) with varying hospital admission probabilities most likely reduces the risk that our results were wholly due to Berkson's bias [16]. 2.3. Definition of MS MS was the main exposure variable. MS was defined as the presence of an ICD-9-CM code beginning with 340 in any of the secondary discharge diagnosis fields. Patients who did not have a code beginning with 340 in any of their secondary discharge diagnosis fields were considered to be free of MS. 2.4. Data analysis Data were analyzed using SAS 9.3 software (SAS Institute, Inc., Cary, North Carolina). Initial analyses involved the creation of contingency tables with a significance level of 0.05. Unadjusted and adjusted odds ratios (OR), 95% confidence intervals (CI), and P values were calculated from unconditional logistic regression models. ORs for the association between MS and BC were adjusted for the patient's age (seven age groups modeled using six indicator variables), the patient's raceethnicity (four groups), and the patient's health insurance information (3 groups). The original health insurance variable found in our dataset has over 20 possible response values. We collapsed these categories and created a new health insurance variable with the following three groups: (1) self-pay or indigent (combined together, see Anderson, 2007 [17]), (2) Medicaid, and (3) cases that are not in either group (1) or (2). Parity is another possible confounding factor in any analysis in which the outcome is BC and MS since nulliparous women have a higher risk of developing BC [18] and nulliparity is more frequent among MS patients [19]. We sought to control for parity indirectly by adjusting
Table 1 Characteristics of the study sample. Female breast cancer (BC) cases were compared with female type II diabetic controls (DC) and female open wound (OW) controls. Controls did not have a secondary discharge diagnosis of BC. The patients were discharged throughout Texas between 2004 and 2007 and found in the Public Use Data File. Variable
BC cases
DC
P
OW controls
BC cases vs. DC
Age (years) 20–29 30–39 40–49 50–59 60–69 70–79 ≥80 Race-ethnicity Black non-Hispanic White Hispanic White non-Hispanic Other (Asian, Native American, etc.) Health insurance Self-pay, indigent Medicaid Other Has multiple sclerosis
N = 21,536
N = 54,141
Number (%)
Number (%)
BC cases vs. OW Controls N = 5440 Number (%)
b0.0001 156 (0.72) 1348 (6.3) 4394 (20.4) 5498 (25.5) 4509 (20.9) 3490 (16.2) 2141 (9.9)
1443 (2.7) 3729 (6.9) 7959 (14.7) 12,212 (22.6) 11,198 (20.7) 10,369 (19.2) 7231 (13.4)
2812 (13.1) 1355 (6.3) 14,015 (65.1) 3354 (15.6)
12,955 (23.9) 7550 (14.0) 19,497 (36.0) 14,139 (26.1)
1574 (7.3) 1309 (6.1) 18,653 (86.6) 37 (0.17)
6799 (12.6) 6582 (12.2) 40,760 (75.3) 103 (0.19)
P
b0.0001 811 (14.9) 721 (13.3) 888 (16.3) 783 (14.4) 540 (9.9) 634 (11.7) 1063 (19.5)
b0.0001
b0.0001 687 (12.6) 511 (9.4) 3285 (60.4) 957 (17.6)
b0.0001
0.59
b0.0001 957 (17.6) 395 (7.3) 4088 (75.2) 17 (0.31)
0.04
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
for the patient's age since further analysis of the study by Mulla et al. demonstrated that nulliparity at the time of presentation to a labor and delivery unit among women delivering at a single hospital declined steadily with increasing age of the woman, and hence a woman's age can be a proxy measure for her parity [20]. We also analyzed the frequency distribution of the principal procedure field among BC cases. If an adjusted MS OR differed by 10 or more percent from the crude MS OR, then confounding was considered to be present [21].
3
63% less likely than OW controls to have MS (adjusted OR = 0.37, P = 0.001). We attempted to conduct a second case–control study among men (ICD9 code 175, malignant neoplasm of the male breast; n = 136 with eligible values for patient age, race-ethnicity, and health insurance), but none had a secondary discharge diagnosis code for MS so further analyses were not completed. 4. Discussion
3. Results The flowchart in Fig. 1 presents frequency distribution of the principal procedure code for our BC cases; i.e. the number of cases and controls taken into our analyses. We identified 21,600 female cases with the principal discharge diagnosis of breast cancer (BC). However, 124 patients were removed from the study because of invalid data for age, race/ethnicity or health insurance information. As a result, 21,536 cases were taken into the further analysis. Similar evaluation of data for two control series yielded 54,141 patients with diabetes type II and 5440 patients with open wounds. Characteristics of the study sample are reported in Table 1. A statistically-significant difference in the age distribution was found between BC cases and diabetic controls (DC) (P b 0.0001), and BC cases and open wound (OW) controls (P b 0.0001). The prevalence of MS was 0.17% in BC cases and 0.19% (P = 0.59) and 0.31% (P = 0.04) in diabetic and open wounds controls, respectively. The frequency distribution of the principal procedure that was performed during the hospital stay of the BC patients was also examined. This analysis showed that 77% of the BC cases had undergone surgical procedures for the management of breast cancer. For example, 9741 BC patients had unilateral extended simple mastectomy (ICD-9 CM code 85.43), while 2897 BC patients had unilateral simple mastectomy (ICD-9 CM code 85.41), procedures that are usually performed early after BC diagnosis, suggesting that most of the BC cases that were included in our analysis were likely newly diagnosed BC cases. Unadjusted (crude) and adjusted ORs for the relationship between MS and the outcome of BC are shown in Table 2. Both adjusted ORs differ by ≥10% from their respective unadjusted ORs. After controlling for age, race-ethnicity, and health insurance status, a statistically-significant protective association was detected: BC cases were 45% less likely than DC controls to have MS (adjusted OR = 0.55, P = 0.003), and
Our novel analysis of a population-based hospital dataset found an inverse association between MS and the outcome of BC, suggesting a decreased risk for BC in MS patients, given that in the majority of cases MS precedes BC: the average age of MS diagnosis is 36 years [22] and for BC diagnosis the average age is 61 [23]. Breast cancer diagnosed in women 35 years of age or less accounts for b 2% of all breast cancer cases [24] further supporting our assumption that MS precedes BC in majority of cases. Two other studies also found a decreased prevalence and incidence of BC in MS patients, in US and France, respectively [25,26]. However, conflicting findings abound in the current literature regarding the comorbidity of BC and MS in other countries. For example, some studies found an increased incidence of BC in MS patients in Taiwan or Norway, while other studies found no significant association between MS and BC in Finland, Israel, or Canada (for a detailed review see Marrie et al., 2015 [13]). As discussed in Marrie et al. [13], the literature characterizing the risk of all cancer (including BC) in MS is typified by incongruent study designs and a lack of consistent adjustment for various confounders, and thus can explain the observed variability among different reports. It is also possible that the presence or absence of association between MS and BC in various populations that were analyzed at different times reflects the variability in pharmacological treatments of analyzed MS patients. There are currently at least seven different diseasemodifying treatments for MS (reviewed in Files et al., 2015 [27]), most of which become available only recently. Our study adjusted for age, race and health insurance, and used a comprehensive statewide hospital discharge database obtained from the THCIC, rather than analyzing data from a single institution or a limited geographical area. In 2007, quarterly figures indicated that on average 80% of the state licensed hospitals in Texas were required to report their data to the THCIC and of this group 99% did report (Personal communication, THCIC staff, Austin, Texas). Furthermore, Texas is a state
Fig. 1. Numbers of individuals in our study sample. *Cases were deemed invalid if the values for age, and/or race/ethnicity and/or health insurance were ineligible, missing, or nonsensical.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
4
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
Table 2 Unadjusted and adjusted odds ratios (OR), 95% confidence intervals (CI), and P values for the association between multiple sclerosis (MS), present vs. absent, and the outcome of breast cancer (BC) in women discharged throughout Texas. BC cases (N = 21,536) were compared to female type II diabetic controls (DC) (N = 54,141) and female open wound (OW) controls (N = 5440). controls did not have a secondary discharge diagnosis of BC. Type of OR
DC MS OR (95% CI)
Unadjusted Adjusted for the patient's age, race-ethnicity, and health insurance
OW controls P
0.91 0.60 (0.62–1.32) 0.55 0.003 (0.37–0.81)
MS OR (95% CI)
P
0.55 0.04 (0.31–0.97) 0.37 0.001 (0.21–0.66)
that is large both in terms of population (estimated at 22,778,123 in 2005; United States Census Bureau) and size (land area in 2010 was approximately 261,232 mile2; Texas QuickFacts; U.S. Census bureau: 2015) indicating that our study involved a large geographical area and a large number of patients: 21,536 cases and 59,581 controls. Although diagnoses found in hospital discharge datasets may suffer from miscoding, a review of the validity of the ICD coding of neurological conditions in administrative health data by St. Germaine Smith et al. (2012) found that the coding of MS was accurate (sensitivity ranging from 85 to 92.4%, and a positive predictive value of 74.5 to 92.7%) [28]. Another possible, but unlikely limitation of our study stems from the fact that the Texas Inpatient Public Use Data File (PUDF) does not include a unique patient identifier, and hence there is a chance that multiple records were included in our sample for one or more patients. It is possible that patients were transferred between two hospitals both of whom report to the PUDF. However, our analysis showed that only 1.6% of our BC cases, 4.5% of our diabetic controls, and 6.8% of our OW controls had an unknown value for their source of admission variable or had been identified as transfers from other hospitals. Furthermore, it is possible that the initial admission was at a smaller facility that is not required to report their patient discharge data to the PUDF and hence only one record for the patient was included in our sample. Therefore, this limitation of de-identified hospital inpatient discharge dataset unlikely affected the result of our study. It is also possible that an individual was discharged to their home and was re-admitted to the same facility due to a complication; however, in this situation it is likely that the complication, e.g., an infection, rather than BC, would be noted in the principal discharge diagnosis field. Potential interpretations of the inverse association between BC and MS obtained in our study include the possibility that MS treatments of analyzed patients decreased the risk for developing BC, conceivable common biological mechanisms, and/or that the biased selection of hospitalized cases and controls influenced the outcome in our study. Although we could not analyze MS treatments (data were not available), it is possible that MS drugs influence the development of BC, as suggested in some studies. For example, an Israeli study reports that interferon beta (IFNβ), which has been approved for the treatment of MS since 1993 (and thus it was likely used by MS patients in our study) lowers the risk of BC in female patients [29]. Furthermore, recent studies also demonstrate that IFNβ has an import role in the BC pathology [30], and IFNβ is now considered as a possible intervention for BC [31]. Despite evident clinical relevance, epidemiological studies addressing the effect of MS treatments on BC are scarce and inconclusive [32]. In kind, immunomodulatory treatments that are potentially beneficial for both MS and BC may imply common biological mechanisms underlying both diseases. Nevertheless, similar immune reactions [33] and the interplay of sex-specific factors or the genetic background [34] linking MS and BC cannot fully explain a potentially protective effect of MS on BC. We attempted to shed light on the possible contribution of the sex-specific factors to the risk of BC in MS by also analyzing male population in our data base. However, given that BC is 100 times less common in men [35], we identified only 136 men with BC, and
since MS is more than twice as common in women than in men [36], we expectedly did not identify any men with both BC and MS. Although the comparison between BC risks in men and women with MS has not yet been reported, Marrie et al. have shown that the percent of selfreported BC in men with MS (3%) is higher than in women with MS (2.3%), indicating a possibility that sex-specific factors contribute to the risk of BC in MS patients [12]. Any hospital-based case–control study in which both the exposure and the outcome are diseases may suffer from Berkson's bias, [16,37]. We attempted to minimize the possibility that our results were wholly due to Berkson's bias by using two control conditions, one acute and one chronic, which may have different hospital admission probabilities. The use of multiple control series can be viewed as a form of sensitivity analysis. Our estimation of the strength of the association between MS and BC was largely insensitive to the choice of the control disease/condition: both of the adjusted exposure ORs (Table 2) indicated the presence of a strong, protective relationship. Furthermore, because the use of prevalent cases may lead to biased ORs due to the effect of selective survival, we confirmed that the majority of our BC cases were newly diagnosed (incident), thereby reducing the possibility of selective survival bias. Even though cases in our study closely matched the age distribution of BC in general population, the race/ethnicity distribution did not, with 2/3 of analyzed women being white non-Hispanics, thus reflecting unbalanced admission probabilities and/or unequal BC incidence (at least for different races/ethnicities). Therefore, it is possible that marked underrepresentation of black women in our study affected the overall association between MS and BC, as black women have higher incidence of BC than white women [35]. This implication raises an interesting possibility that the risk for BC in MS patients depends on the race/ethnicity, which remains to be investigated further. Our study presented here is the only recent assessment of the association between MS and BC in a relatively large area of the US. It is worth noting that only two reports [11,25] that analyzed MS and BC association in the US preceded ours by a few decades; i.e. analyzed data were collected from 1975 to 1984, or in 1989, respectively. In addition, those two studies investigated more restricted populations, i.e. in the Olmsted County (Minnesota), or only Medicare patients older than 65. Our results suggest a protective effect of MS in the hospital-based population of BC patients, in agreement with earlier reports [25,26]. Given the clinical implication of such association, both more comprehensive epidemiological investigations of the USA population, and studies analyzing biological underpinnings of potentially protective effects of MS are sorely needed.
Conflict of interest None exists.
Acknowledgments This work was in part supported by the 2014 Paul L. Foster School of Medicine Individual Investigator-Initiated Seed Grant Program (P.I. ON). The authors acknowledge the use of the Texas Hospital Inpatient Discharge Public Use Data File, quarters 1 through 4, for the years 2004 through 2007, Texas Department of State Health Services, Center for Health Statistics-Texas Health Care Information Collection, Austin, Texas, USA.
Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jns.2015.06.033.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
P.W. O'Malley et al. / Journal of the Neurological Sciences xxx (2015) xxx–xxx
References [1] M.J. Tullman, Overview of the epidemiology, diagnosis, and disease progression associated with multiple sclerosis, Am. J. Manag. Care 19 (2 Suppl.) (2013) S15–S20. [2] L. Moroni, I. Bianchi, A. Lleo, Geoepidemiology, gender and autoimmune disease, Autoimmun. Rev. 11 (6–7) (2012) A386–A392. [3] M. Kipp, S. Amor, R. Krauth, C. Beyer, Multiple sclerosis: neuroprotective alliance of estrogen-progesterone and gender, Front. Neuroendocrinol. 33 (1) (2012) 1–16. [4] R.D. Spence, R.R. Voskuhl, Neuroprotective effects of estrogens and androgens in CNS inflammation and neurodegeneration, Front. Neuroendocrinol. 33 (1) (2012) 105–115. [5] B.F. Bebo Jr., E. Zelinka-Vincent, G. Adamus, D. Amundson, A.A. Vandenbark, H. Offner, Gonadal hormones influence the immune response to PLP 139–151 and the clinical course of relapsing experimental autoimmune encephalomyelitis, J. Neuroimmunol. 84 (2) (1998) 122–130. [6] Group EHBCC, Body mass index, serum sex hormones, and breast cancer risk in postmenopausal women, J. Natl. Cancer Inst. 95 (16) (2003) 1218–1226. [7] R. Kaaks, K. Tikk, D. Sookthai, H. Schock, T. Johnson, A. Tjønneland, et al., Premenopausal serum sex hormone levels in relation to breast cancer risk, overall and by hormone receptor status—results from the EPIC cohort, Int. J. Cancer 134 (8) (2014) 1947–1957. [8] N. Heldring, A. Pike, S. Andersson, J. Matthews, G. Cheng, J. Hartman, et al., Estrogen receptors: how do they signal and what are their targets, Physiol. Rev. 87 (3) (2007) 905–931. [9] S. Gobbi, A. Rampa, F. Belluti, A. Bisi, Nonsteroidal aromatase inhibitors for the treatment of breast cancer: an update, Anti Cancer Agents Med. Chem. 14 (1) (2014) 54–65. [10] C. Holzmann, I. Bauer, P. Meyer, Co-occurrence of multiple sclerosis and cancer in a BRCA1 positive family, Eur. J. Med. Genet. 56 (10) (2013) 577–579. [11] D.R. Wynn, M. Rodriguez, W.M. O'Fallon, L.T. Kurland, A reappraisal of the epidemiology of multiple sclerosis in Olmsted County, Minnesota, Neurology 40 (5) (1990) 780–786. [12] R. Marrie, R. Horwitz, G. Cutter, T. Tyry, D. Campagnolo, T. Vollmer, Comorbidity, socioeconomic status and multiple sclerosis, Mult. Scler. 14 (8) (2008) 1091–1098. [13] R.A. Marrie, N. Reider, J. Cohen, O. Stuve, M. Trojano, P.S. Sorensen, et al., A systematic review of the incidence and prevalence of cancer in multiple sclerosis, Mult. Scler. 21 (3) (2015) 294–304. [14] P. Tettey, S. Simpson Jr., B.V. Taylor, I.A.F. van der Mei, The co-occurrence of multiple sclerosis and type 1 diabetes: shared aetiologic features and clinical implication for MS aetiology, J. Neurol. Sci. 348 (1–2) (2015) 126–131. [15] G.S. Watson, S. Craft, Insulin resistance, inflammation, and cognition in Alzheimer's Disease: lessons for multiple sclerosis, J. Neurol. Sci. 245 (1–2) (2006) 21–33. [16] J. Schwartzbaum, A. Ahlbom, M. Feychting, Berkson's bias reviewed, Eur. J. Epidemiol. 18 (12) (2003) 1109–1112. [17] G.F. Anderson, From ‘Soak The Rich’ To ‘Soak The Poor’: recent trends in hospital pricing, Health Aff. 26 (3) (2007) 780–789. [18] J. Russo, G.A. Balogh, I.H. Russo, Full-term pregnancy induces a specific genomic signature in the human breast, Cancer Epidemiol. Biomark. Prev. 17 (1) (2008) 51–66. [19] P. Cavalla, V. Rovei, S. Masera, M. Vercellino, M. Massobrio, R. Mutani, et al., Fertility in patients with multiple sclerosis: current knowledge and future perspectives, Neurol. Sci. 27 (4) (2006) 231–239.
5
[20] Z.D. Mulla, T. Carrillo, R. Kalamegham, L.L. Hernandez, E. Portugal, B.S. Nuwayhid, Is maternal colonization with group B streptococci a risk factor for preeclampsia? J. Reprod. Med. 60 (3–4) (2015) 117–126. [21] S. Greenland, Modeling and variable selection in epidemiologic analysis, Am. J. Public Health 79 (3) (1989) 340–349. [22] M.L. Sumelahti, E. Pukkala, M. Hakama, Cancer incidence in multiple sclerosis: a 35year follow-up, Neuroepidemiology 23 (5) (2004) 224–227. [23] N. Howlander, A. Noone, M. Krapcho, J. Garshell, D. Miller, S. Altekruse, SEER Cancer Statistics Review, 1975–2011, National Cancer Institute, Bethesda, MD, 2014. [24] S. Bayraktar, L. Amendola, A.M. Gutierrez-Barrera, S.S. Hashmi, C. Amos, M. Gambello, Clinicopathologic characteristics of breast cancer in BRCA-carriers and non-carriers in women 35 years of age or less, The Breast 23 (6) (2014) 770–774. [25] S.T. Fleming, R.L. Blake Jr., Patterns of comorbidity in elderly patients with multiple sclerosis, J. Clin. Epidemiol. 47 (10) (1994) 1127–1132. [26] C. Lebrun, M. Debouverie, P. Vermersch, P. Clavelou, L. Rumbach, J. de Seze, et al., Cancer risk and impact of disease-modifying treatments in patients with multiple sclerosis, Mult. Scler. 14 (3) (2008) 399–405. [27] D.K. Files, T. Jausurawong, R. Katrajian, R. Danoff, Multiple sclerosis, Prim. Care 42 (2) (2015) 159–175. [28] C. St. Germaine-Smith, A. Metcalfe, T. Pringsheim, J.I. Roberts, C.A. Beck, B.R. Hemmelgarn, et al., Recommendations for optimal ICD codes to study neurologic conditions: a systematic review, Neurology 79 (10) (2012) 1049–1055. [29] A. Achiron, Y. Barak, M. Gail, M. Mandel, D. Pee, R. Ayyagari, et al., Cancer incidence in multiple sclerosis and effects of immunomodulatory treatments, Breast Cancer Res. Treat. 89 (3) (2005) 265–270. [30] H. Xu, J. Xian, E. Vire, S. McKinney, V. Wei, J. Wong, et al., Up-regulation of the interferon-related genes in BRCA2 knockout epithelial cells, J. Pathol. 234 (3) (2014) 386–397. [31] A. Carpi, A. Nicolini, A. Antonelli, P. Ferrari, G. Rossi, Cytokines in the management of high risk or advanced breast cancer: an update and expectation, Curr. Cancer Drug Targets 9 (8) (2009) 888–903. [32] E. Kingwell, C. Evans, F. Zhu, J. Oger, S. Hashimoto, H. Tremlett, Assessment of cancer risk with β-interferon treatment for multiple sclerosis, J Neurol Neurosurg Psychiatry 85 (10) (2014) 1092–1102. [33] M. Anagnostouli, G. Anagnostoulis, S. Katsavos, M. Panagiotou, E. Kararizou, P. Davaki, HLA-DRB1*15:01 and Epstein–Barr virus in a multiple sclerosis patient with psoriasis, nasopharyngeal and breast cancers. Lessons for possible hidden links for autoimmunity and cancer, J. Neurol. Sci. 339 (1–2) (2014) 26–31. [34] J.J. Chen, F. Costello, R.H. Kardon, Sex disparities in neuro-ophthalmologic disorders, Curr. Eye Res. 40 (2) (2015) 247–265. [35] B.A. Kohler, R.L. Sherman, N. Howlader, A. Jemal, A.B. Ryerson, K.A. Henry, et al., Annual report to the nation on the status of cancer, 1975–2011, featuring incidence of breast cancer subtypes by race/ethnicity, poverty, and state, J. Natl. Cancer Inst. 107 (6) (2015). [36] A. Alonso, M.A. Hernan, Temporal trends in the incidence of multiple sclerosis: a systematic review, Neurology 71 (2) (2008) 129–135. [37] J. Berkson, Limitations of the application of fourfold table analysis to hospital data, Biometrics 2 (3) (1946) 47–53.
Please cite this article as: P.W. O'Malley, et al., Multiple sclerosis and breast cancer, J Neurol Sci (2015), http://dx.doi.org/10.1016/ j.jns.2015.06.033
