You are here

Article Text

Article menu

Download PDFPDF

Social inequalities in health
Adverse socioeconomic position during the life course is associated with multiple sclerosis
  1. Farren B S Briggs1,
  2. Brigid S Acuña2,
  3. Ling Shen2,
  4. Kalliope H Bellesis2,
  5. Patricia P Ramsay1,
  6. Hong Quach1,
  7. Allan Bernstein3,
  8. Catherine Schaefer2,
  9. Lisa F Barcellos1,2
  1. 1Division of Epidemiology, Genetic Epidemiology and Genomics Laboratory, School of Public Health, University of California, Berkeley, California, USA
  2. 2Kaiser Permanente Division of Research, Oakland, California, USA
  3. 3Palm Drive Hospital, Sebastopol, California, USA
  1. Correspondence to Dr Lisa F Barcellos, Division of Epidemiology, Genetic Epidemiology and Genomics Lab, School of Public Health, 308D Stanley Hall, University of California, Berkeley, CA 94720-3220, USA; lbarcellos{at}genepi.berkeley.edu

Abstract

Background Adverse socioeconomic position (SEP) in childhood and adulthood is associated with a proinflammatory phenotype, and therefore an important exposure to consider for multiple sclerosis (MS), a complex neuroinflammatory autoimmune disease. The objective was to determine whether SEP over the life course confers increased susceptibility to MS.

Methods 1643 white, non-Hispanic MS case and control members recruited from the Kaiser Permanente Medical Care Plan, Northern California Region, for which comprehensive genetic, clinical and environmental exposure data have been collected were studied. Logistic regression models investigated measures of childhood and adulthood SEP, and accounted for effects due to established MS risk factors, including HLA-DRB1*15:01 allele carrier status, smoking history, history of infectious mononucleosis, family history of MS and body size.

Results Multiple measures of childhood and adulthood SEP were significantly associated with risk of MS, including parents renting versus owning a home at age 10: OR=1.48, 95% CI 1.09 to 2.02, p=0.013; less than a college education versus at least a college education based on parental household: OR=1.28, 95% CI 1.01 to 1.63, p=0.041; low versus high life course SEP: OR=1.50, 95% CI 1.09 to 2.05, p=0.012; and low versus high social mobility: OR=1.74, 95% CI 1.27 to 2.39, p=5.7×10−4.

Conclusions Results derived from a population-representative case–control study provide support for the role of adverse SEP in MS susceptibility and add to the growing evidence linking lower SEP to poorer health outcomes. Both genetic and environmental contributions to chronic conditions are important and must be characterised to fully understand MS aetiology.

  • Multiple Sclerosis
  • Epidemiology
  • Lifecourse / Childhood Circumstances
  • Social Epidemiology
  • Social and life-course epidemiology

https://doi.org/10.1136/jech-2013-203184

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Multiple sclerosis (MS) is a complex autoimmune disease with two distinct but overlapping neuropathological phases, inflammatory and neurodegenerative.1 Susceptibility to MS is multifactorial, involving both genetic and environmental components. Progress has been made on the characterisation of several MS risk factors (eg, HLA-DRB1*15:01 and other genetic variants, exposure to tobacco smoke, Epstein–Barr virus (EBV) infection, residential latitude (which may capture sunlight exposure and related vitamin D levels) and obesity); however, the underlying aetiology remains unexplained.2–7

    Socioeconomic position (SEP) has emerged as an important ‘exposure’ with significant health consequences, where individuals with lower SEP are more likely to suffer increased morbidity and mortality across a wide range of diseases.8 Results from the 1958 British Birth Cohort and The Framingham Offspring Study demonstrated strong associations between low SEP across the life course and elevated inflammatory markers measured in serum.9 ,10 In addition, adverse SEP in childhood has been associated with a proinflammatory phenotype.11 ,12 Therefore, SEP is an important exposure to consider for MS, a complex neuroinflammatory autoimmune disease. Thus far, the impact of SEP on MS risk is unclear, and previous studies have generated conflicting results.13–17 Most recently, an inverse relationship between higher education and MS risk was observed in the Norwegian Registry of Employers and Employees; a rate ratio (RR) of 0.48 for workers with a graduate degree compared with workers who only completed elementary school was reported.18 A similar relationship was observed for childhood SEP and risk of MS in a national Danish cohort.19 Individuals had a reduced rate of MS in later life (RR=0.86), if at age 15 years, their maternal education was greater than secondary level compared with individuals whose highest maternal education was at the primary school level. To date, no study has simultaneously investigated the relationship between childhood and adulthood SEP and MS risk. It is important to assess exposures in childhood and adulthood as both timeframes are critical periods for MS risk.4–7 ,17 ,19–22 SEP can be considered as a chronic exogenous stressor linked to allostatic load and increased vulnerability to negative health outcomes resulting from the dysregulation of multiple biological pathways, therefore considering exposure over the life course is necessary.23 ,24

    We comprehensively investigated the relationship between multiple adverse SEP frameworks and MS susceptibility in a large, population-representative case–control sample. Adverse childhood and adulthood SEP and their cumulative effect in MS were studied, for the first time, in models accounting for known heritable factors and established environmental risk factors.

    Methods

    Study population

    MS cases and controls were identified among members of Kaiser Permanente Medical Care Plan, Northern California Region (KPNC), using electronic medical records (see online supplementary materials). KPNC is an integrated, private, health services delivery system with a membership of 3.2 million that comprises about 30% of the population of a 22 county service area. It is the largest healthcare provider in Northern California, and membership is most commonly linked to employment. Comparisons with the general population have shown the membership is objectively representative, although persons in impoverished neighbourhoods (where >20% of individuals are below the federal US poverty line) are underrepresented25 (see online supplementary materials).

    Briefly, MS cases were required to have one or more outpatient MS diagnoses by a neurologist (multiple sclerosis, ICD9 code 340.xx; 95% had at least two MS diagnoses by a neurologist), self-identified white (non-Hispanic) race/ethnicity, age 18–69 years, and KPNC membership at initial contact. MS cases for study were approved as potential study participants by their neurologist and were required to meet well-established diagnostic criteria.26 ,27 Electronic medical record review, including MRI reports, and patient screening were also used to confirm a diagnosis of MS. Controls were randomly selected from current KPNC members who did not have a MS diagnosis or related condition (optic neuritis, transverse myelitis or demyelination disease; ICD9 codes: 340, 341.0, 341.1, 341.2, 341.20, 341.21, 341.22, 341.8, 341.9, 377.3, 377.30, 377.39 and 328.82) confirmed through electronic medical records. Controls were matched to cases on gender, birth date (± 1 year), self-identified race/ethnicity and ZIP code of the case residence. At the time of the data freeze (December 2010), the average participation rate was 58% for controls and 79% for cases; however, KPNC length of membership did not differ between the two groups (see online supplementary table 1). Because all cases were not yet matched to a control, unmatched analyses adjusted for the matching variables were conducted.

    Participants were mailed a consent form to be returned by mail and a biospecimen collection kit, to be completed at a KPNC clinic laboratory or to be returned by mail. Onset was year of first self-reported symptom; environmental exposure, sociodemographic and additional clinical and other health related data were also collected for all study participants via a series of standardised questions through computer-assisted telephone interview by trained staff interviewers (see online supplementary materials). Study participants had the choice to decline answering any question or respond ‘Don't know’. Responses that were declined or ‘Don't know’ were coded as missing. On average, no variable had more that 5% responses classified as missing. The study protocol was approved by the Institutional Review Boards of KPNC and the University of California, Berkeley.

    Life course SEP frameworks

    Three models have been developed to conceptualise developmental/life course socioeconomic experiences as determinants of adult health: (1) the sensitive period framework emphasises that exposure to low SEP at specific periods may have greater and/or more lasting influences on adult health; (2) the accumulation of risk framework considers the cumulative insult of SEP over the life course; and, lastly, (3) the social mobility framework represents the stability or change in SEP between generations, such that trajectories or transitions in social states may affect physiological and psychological states.28 These frameworks are not mutually exclusive but allow for a comprehensive investigation of the window and length of exposure to adverse SEP conditions.10

    The sensitive period framework evaluated individual-level SEP and MS risk. Childhood SEP was investigated using parental household education level and parental home ownership. Participants reported the highest education level attained by each parent on an eight-point scale: ‘none’ (7); ‘grade school only (1–8)’ (6); ‘some high school or not a high school graduate’ (5); ‘high school graduate or GED (General Educational Development test)’ (4); ‘some college or technical/trade/vocation school or associate's degree’ (3); ‘bachelor's degree’ (2); ‘master's degree’ (1); and ‘doctoral degree’ (0). Parental household education level was the highest education level attained by either parent or the highest education level of one parent, if the other parent's information was not reported (see online supplementary figure 1). We were interested in the effect of adverse SEP on MS risk, therefore parental household education level was considered as a continuous (where ‘doctoral degree’=0 and ‘none’=7) and a binary predictor, dichotomised at the collegiate level (≥ bachelor's degree=0 and <bachelor's degree=1) and these analyses were restricted to MS cases with onset ≥18 years of age. Participants were asked to recall whether their parents owned or rented their homes when the participant was 10 years of age (parents owned=0 and parents rented=1). Adulthood SEP was investigated using the participant's educational attainment, reported on the eight-point scale outlined (see online supplementary figure 2). Participant's education level was also considered as a continuous and a binary predictor, dichotomised at the collegiate level.

    The accumulation of risk framework was assessed using a cumulative score that included household education and participant's education.10 A score for each SEP measure was categorised separately and then summed (household education: high school diploma or less=0; more than a high school diploma and less than a college degree=1; college degree=2; postgraduate degree=3; similarly for participant's education). The summed SEP score (range 0–6) was classified into low (score=0–1), medium (score=2–3) and high (score=4–6) cumulative SEP exposures. The social mobility framework was investigated using a categorical variable that captured SEP trajectories from childhood to adulthood based on collegiate education levels: low–low (household and participant's education < bachelor's degree); low–high (household education < bachelor's degree, participant's education ≥ college degree); high–low (household education ≥ college degree, participant's education < college degree); and high–high (household and participant's education ≥ college degree) trajectories.10 To minimise the potential for reverse causality of MS on educational-level analyses of adulthood SEP (sensitive period framework), the accumulation of risk framework and the social mobility framework were restricted to MS cases with onset age ≥28 years to allow for at least 10 years from expected high school graduation to achieve the maximum years of education not likely to be influenced by the disease process.

    Statistical analyses

    Prior to all analyses, genetically related individuals and genetic outliers were removed (see online supplementary materials). Pearson's pairwise correlation assessed the dependence between all SEP variables. Logistic regression models estimated crude and adjusted ORs and 95% CI. Models were adjusted for conventional variables and risk factors relevant to MS: year of birth, gender, history of cigarette smoking (ever or never; where ever was defined as smoking at least one cigarette a day for at least 1 month), history of infectious mononucleosis (ever or never), history of MS among first-degree relatives (any or none) and body size at ages 10 and 20. Study participants were asked to recall self-perceived body size at ages 10 and 20 from one of four categories: ‘underweight’, ‘just about right’, ‘little overweight’ or ‘very overweight’. Weight in both childhood and adulthood have been shown to be independently associated with MS risk.6 ,7 Therefore, body size at age 10 was used in models where MS cases had onset age ≥18 years, and body size at age 20 was used in models where MS cases had onset age ≥28 years. Recent evidence shows social class may be associated with vitamin D levels in a contemporary UK birth cohort (1991–1992).29 Latitude of birth (see online supplementary materials), latitude of residence at age 10 (see online supplementary materials), vitamin D supplementation at age 10 (ever or never) or at age 28 variables were included (table 1); however, the distribution of each did not differ between cases and controls, and they were not included in final models. Models investigating the sensitive period framework were also adjusted for other SEP variables; for example, the parents renting a home at age 10, parental household education and less than a college education based on parental household models were adjusted for continuous measure of participant's education and median household income for current census block (see online supplementary materials); while participant's education and less than a college education in adulthood models were adjusted for parental household education and median household income for current census block. Inclusion of multiple SEP predictors in the model did not inflate the variance, allowing for concurrent adjustment (variance inflation factor <5; data not shown). Cumulative SEP and social mobility scores were treated as categorical predictors; however, continuous variables were used to determine ptrend. Models were also adjusted for carrier status of HLA-DRB1*15:01 alleles and the known non-HLA genetic component based on 52 loci associated with MS through recent GWAS using a weighted genetic risk score (wGRS) as previously described (see online supplementary materials).3 ,30 We adjusted our models for established genetic and exogenous risk exposures as most are proinflammatory, suggestive of a proinflammatory phenotype, or may affect immune response. Adverse SEP has been associated with a proinflammatory phenotype. Because serum measures of inflammation over the life course were not available for study participants, we instead accounted for the variance explained by established risk factors to identify the independent effects in models due to adverse SEP. Missing data were included using an indicator variable for categorical variables and were non-informative (data not shown). The study was well powered to detect modest effects (see online supplementary materials). The log-likelihood ratio statistic compared all SEP models with a baseline model with all other covariates using MS cases with onset age ≥28 years and healthy controls; Akaike information criterion and Bayesian information criterion were also assessed (see online supplementary table 2). All analyses were conducted in STATA V.11.2 (StataCorp, Texas, USA) using the pwcorr, logit, lrtest and estat ic commands.

    View this table:
    Table 1

    Characteristics of non-Hispanic white MS case–control study participants*

    Results

    There were 1023 MS cases with onset age ≥18 years and 699 MS cases had onset age ≥28 years (68% of cases), and 620 healthy controls. As expected, MS cases were more likely to report ever having infectious mononucleosis, a history of tobacco smoking, a family history of MS and increased body size at ages 10 and 28 compared with controls (table 1). Latitude of birth and latitude of residence at age 10 were similar, regardless of disease status. Vitamin D supplementation at ages 10 and 28 did not significantly differ, though the cases were less likely to have been on supplementation. Cases were significantly more likely to be carriers of HLA-DRB1*15:01, and had a greater wGRS, thus possessed more non-HLA risk variants compared with controls. The median household income for current census block in case and control groups was similar, though slightly higher among controls. There was strong positive correlation between paternal, maternal and household education variables, and a modest association with the participant's education; however, no evidence for correlation between any education variables or parental home ownership at age 10 and the median household income of the census block was observed (table 2). Given that the education, household and median household income of the census block variables were not correlated, we assumed education and household variables did not capture all ‘current neighbourhood’ characteristics.

    View this table:
    Table 2

    Pearson's pairwise correlation among SEP variables in combined dataset of MS cases and controls*

    Under the sensitive period framework, the influence of childhood SEP on MS risk was investigated using parental household education and parental home ownership status at age 10. Crude and adjusted logistic regression models demonstrated similar results for the three childhood SEP measures (table 3). Increased MS risk was associated with renting at age 10. Multivariable logistic regression models adjusted for (1) conventional covariates (Model 1); (2) conventional covariates and other SEP variables, including participant's education, and median household income of census block (Model 2); and, lastly, (3) conventional covariates, other SEP variables including, and genetic risk factors (HLA-DRB1*15:01, wGRS) (Model 3). Results show consistent associations between all childhood SEP predictors and MS in all models (table 3); MS cases were ∼50% more likely to live at age 10 with parents who did not own a home (ORModel 1=1.56, 95% CI 1.16 to 2.09; ORModel 2=1.47, 95% CI 1.10 to 1.98; ORModel 3=1.48, 95% CI 1.09 to 2.02). Lower parental household education was also associated with MS (table 3).

    View this table:
    Table 3

    Results from association testing of adverse childhood socioeconomic position (SEP)*and adulthood SEP† (sensitive period framework) with susceptibility to MS

    Results from analyses of adulthood SEP (participant's education) showed MS susceptibility is associated with lower education levels (table 3). Having less than a college degree was associated with increased risk of MS (ORModel 1=1.47, 95% CI 1.17 to 1.85; ORModel 2=1.32, 95% CI 1.03 to 1.69; ORModel 3=1.30, 95% CI 1.00 to 1.69). Estimates did not change when adjusting for the genetic component of MS (Model 3). MS cases were more likely to have both adverse childhood and adulthood SEP exposures compared with controls, even after accounting for known genetic and environmental risk factors.

    Results from the accumulation of risk framework analyses demonstrated strong evidence for association between low and medium cumulative SEP scores and MS risk (table 4). Individuals with a medium cumulative SEP score were 40% more likely to have MS compared with those with a high cumulative SEP score (ORadjusted=1.40, 95% CI 1.05 to 1.86). Individuals with the lowest cumulative SEP score were 50% more likely to have MS compared with those with a high cumulative SEP score (ORadjusted=1.50, 95% CI 1.09 to 2.05). Accumulation of adverse SEP and MS risk showed significant evidence for a trend relationship (p=0.011).

    View this table:
    Table 4

    Marginal and adjusted ORs demonstrating the association between adverse socioeconomic position under the accumulation of risk and social mobility frameworks with increased susceptibility to MS*

    Under the social mobility framework, MS risk was strongly associated with adverse SEP conditions, similar to results under the other frameworks (table 4). Individuals with both high SEP in childhood and adulthood (high–high) were used as the reference category. The strongest associations were observed between MS and both upwardly mobile SEP trajectory (low–high) and stable low SEP trajectory (low–low) variables; for both comparisons, a low childhood SEP was present (low–high: ORadjusted=1.52, 95% CI 1.07 to 2.15; and low–low: ORadjusted=1.74, 95% CI 1.27 to 2.39). Overall, social mobility and MS risk showed significant evidence for a trend relationship (p=9.8×10−4).

    Discussion

    SEP is predictive of chronic disease and long-term health outcomes in adults. To investigate this possibility in MS, we used comprehensive exposure and genetic data collected from 1643 white, non-Hispanic MS cases and controls. Significant associations between adverse SEP and MS were observed using multiple frameworks, which have not been previously investigated in MS, but are consistent with recent findings.18 ,19 This is the first study to illustrate independent effects of adverse childhood and adulthood SEP in MS, even after accounting for known heritable factors, other established environmental risk factors, family history of MS and body size at two time points. Results add to the growing evidence that MS is a very complex, multifactorial disease. As medical care moves to more personalised treatment and prevention strategies, both genetic and environmental contributions, in combination, will be important for prediction of MS onset and related clinical outcomes. Genomic sequence data will play a major role; however, the health outcomes of an individual, including response to drug therapy, are likely to be influenced by experience, education, culture and a wide variety other factors.31

    SEP constructs are multidimensional, capturing environmental (ie, pollution, access to resources, residential characteristics), psychosocial (ie, stress exposures, subjective social status) and behavioural (ie, diet, smoking, physical activity) exposures that have the potential to impact multiple biological systems (ie, inflammatory, neuroendocrine), which may be relevant to the pathoetiology of MS.32 As there is no gold standard for measuring SEP,33 we focused on educational attainment as one individual-level measure since it is a relatively stable variable set in early adulthood (<28 years) and may be representative of life course SEP, as it precedes occupation, income and wealth.34 ,35 Household educational attainment has been suggested as a proxy for childhood SEP.36 Further, another measure of childhood SEP is parental home ownership, which is associated with greater assets and income; both are of primary importance in SEP trajectories.37 Given the multidimensional nature of SEP, investigating multiple measures at different life stages has been recommended.33 In particular, socioeconomic disadvantage experienced in childhood has lasting negative influences on adult health, irrespective of subsequent social transitions.38 ,39 In addition, adult SEP greatly contributes to the cumulative impact of SEP on health.8

    For the sensitive period framework, both childhood and adulthood SEP were significantly associated with MS risk. Given that adverse childhood SEP conditions are more likely to lead to lower education attainment, and therefore lower adulthood SEP,40 demonstrated by moderate correlation within our study population (table 2), we adjusted for the reciprocal SEP variable and current census block data. The median household income of the census block served as an unbiased measure of unquantified SEP constructs. Childhood SEP measures were significantly associated with MS risk, whereas adulthood SEP was modestly significant, in multivariable models (p=0.048). The observed difference is likely due to the smaller sample after restricting to cases with disease onset ≥28 years. In addition, adulthood SEP was weakly correlated with census block data; given that both variables were included in multivariable models, modest collinearity may have contributed. It is important to note that the direction and magnitude of effect, as well as significance for both home ownership and parental household education as SEP predictors in a single model did not change (data not shown), suggesting both act independently to influence MS susceptibility.

    The relevance of both childhood and adulthood SEP on disease risk is emphasised under the accumulation of risk framework. A positive association between increasing level of adverse lifetime SEP conditions and greater MS risk was shown. Similarly, under the social mobility framework, adverse SEP conditions in childhood and adulthood (low–low) were strongly associated with increased MS risk, as were social transitions (low–high, high–low), suggesting that irrespective of when experienced in life, adverse SEP conditions increase MS risk. Both frameworks illustrate a possible gradient, or ‘dose-effect’, on MS risk with respect to adverse lifetime SEP conditions; adverse SEP of early development appears to define health trajectories in later life, regardless of subsequent adult SEP.38 We compared the models from all frameworks among MS cases with onset age ≥28 years and healthy controls using the log-likelihood ratio test as recently proposed.41 Using the same study population, the social mobility model was the most significant compared with a baseline model of the established MS risk factors; this model also had the smallest Akaike information criterion value (see online supplementary table 2). However, it is worth noting that the childhood SEP models may be more relevant to a MS population with onset age ≥18 years. Nonetheless, these findings are consistent with other epidemiological studies that support childhood and adolescence as critical exposure periods for MS risk.17 ,19–22

    The effect of SEP may be mediated through inflammation, which, along with neurodegeneration, is involved in MS pathogenesis.1 In healthy adults from two large cohorts, low SEP across the life course was associated with elevated levels of several inflammatory markers.9 ,10 Adverse childhood SEP may promote an exaggerated proinflammatory phenotype in later life. Adolescent females whose families did not own a home in their childhood had elevated toll-like receptor 4 (TLR4) mRNA and lower glucocorticoid receptor (GR) mRNA levels in peripheral blood mononuclear cells (PBMCs), independent of current SEP within the family.11 PBMC gene expression profiles among adults from low childhood SEP conditions defined by parental occupation in childhood show increased expression of genes with response elements for the proinflammatory nuclear factor kappa B transcription factor.12 Immunological processes may therefore be the principal mechanism through which adverse SEP conditions in childhood and over the life course result in negative health outcomes such as MS in later life. Epidemiological evidence shows the connection between SEP and health outcomes may be mediated by vulnerability and response to stress.42

    This investigation had several strengths. First, a well-powered population-based MS case–control sample with information on both childhood and adulthood SEP was used. Second, we reduced the potential effect of reverse causality, specifically health selection, where the disease affects the exposure, by restricting childhood SEP (sensitive period) analyses to MS cases with onset ≥18 years and all other analyses to cases with onset ≥28 years. As a result, inclusion of any case that might have had (rare) paediatric onset was minimised. Further, maximum parental education could have been achieved any time before the birth of the participant or during the participant's life time, though less likely after the participant was an adult. Third, the three SEP frameworks encompass exposure over the life course and were constructed using specific measures robust to recall bias.34–36 Detailed clinical, environmental and genetic data were available for study participants that were included in regression analyses. Models were adjusted for carrier status of the primary MS susceptibility allele, HLA-DRB1*15:01, and the wGRS, which captured the known genetic component of MS. To the best of our knowledge, this is the first time all risk factors investigated in the current study have been considered in one model. Standardised procedures were in place for all data collection. Each question was asked in the exact same manner by trained interviewer staff. The fact that other variables such as smoke exposure and history of infectious mononucleosis showed similar associations in our case–control study as those reported in cohort studies4 ,5 suggests that recall bias due to case status is not contributing to our findings. Additionally, the study presented here, nested in the larger KPNC longitudinal cohort, also included an unbiased SEP construct, specifically census block data. Finally, the distributions of all SEP measures were similar between all prevalent cases and cases with disease onset of 10 years or less, suggesting that recall bias due to disease duration was a minimum.

    There are also some limitations. The biological relevance of SEP measures investigated here is unknown since education and home ownership have varying implications for earning potential and quality of life across various demographic groups. Also, dissociating the interrelatedness between childhood SEP and adulthood SEP using education is challenging, as an individual's education embodies aspects of social opportunities for education and parental choices and constraints that may influence the individual's socioeconomic conditions in childhood and later life. Further, the KPNC MS cases and controls were recruited from Northern California and the study was restricted to white non-Hispanic participants defined by genetic ancestry; therefore, generalisability to non-white populations or populations from other geographic areas may be limited. It is possible that individuals with chronic conditions may be more likely to remain KPNC members, and therefore the prevalence of chronic conditions in our controls might be higher than the targeted population. However, any effect of retaining people with a chronic condition would be relatively small, given the overall stability of the KPNC population. KPNC has great stability among people who have been health plan members for two or more years. Membership retention over 10 years is 78%, including attrition due to mortality. In fact, the average length of KPNC membership in our MS cases and controls was very similar (see online supplementary table 1), which substantially reduces the potential problem of this source of selection bias in our study. If for some reason controls in our study were more likely to have a chronic condition (ie, coronary heart disease, cancer, etc.), our estimate for association with SEP would be conservative and biased towards the null, given the established SEP association with chronic disease. Furthermore, external validity and generalisability of our findings to members of other health plans should not be impacted. However, the study population may differ from the target population with respect to the prevalence of chronic conditions (which includes uninsured who may have a higher prevalence of chronic diseases); therefore, there could be some effect on generalisability of our findings.

    In summary, adverse SEP was significantly associated with increased MS risk using multiple frameworks; results are in line with previous findings.18 ,19 This is the first study to illustrate independent effects of adverse childhood and adulthood SEP in MS, even after accounting for known heritable factors, and other established environmental risk factors. Results for MS are consistent with other published studies describing the importance of socioeconomic inequalities in both physical and mental health, and with a potential role for inflammation in disease pathogenesis.

    What is already known on this subject

    • Adverse socioeconomic position is associated with a proinflammatory phenotype and has been linked to several chronic health conditions. It is an important exposure to consider for multiple sclerosis. The impact of socioeconomic position on multiple sclerosis risk is unclear, though recent studies suggest that both lower childhood and adulthood socioeconomic position may increase risk; however, it is unknown whether these exposures represent independent associations.

    What this study adds

    • This is the first study to illustrate independent effects of adverse childhood and adulthood socioeconomic position in multiple sclerosis, even after accounting for known heritable factors, other established environmental risk factors, family history of multiple sclerosis and body size at two time points. These results add to the growing evidence that multiple sclerosis is a multifactorial disease where both genetic and environmental (exogenous and psychosocial) contributions, in combination, will be important for prediction of disease onset and other clinical outcomes.

    Acknowledgments

    This work was supported by NIH R01NS049510, NIH R01AI076544 and NIH R01ES017080. FBSB is a National MS Society Postdoctoral Fellow (FG 1847A1/1). We thank MS cases and controls for participating in the study. We also thank Janelle Noble, Julie Lane, Candi Farlice, Adam Boroian, Carol Rabello, Diana Quach, Ellen Mowry, Xiaorong Shao, Emon Elboudwarej, Michaela George and The International MS Genetics Consortium for their contributions to this study. We are very grateful for constructive discussions with Miranda Ritterman Weintraub and Benjamin Goldstein.

    References

    1. ↵
      1. Oksenberg JR,
      2. Barcellos LF
      . Multiple sclerosis genetics: leaving no stone unturned. Genes Immun 2005;6:375–87.
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Hafler DA,
      2. Compston A,
      3. Sawcer S,
      4. et al
      . Risk alleles for multiple sclerosis identified by a genomewide study. N Engl J Med 2007;357:851–62.
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Sawcer S,
      2. Hellenthal G,
      3. Pirinen M,
      4. et al
      . Genetic risk and a primary role for cell-mediated immune mechanisms in multiple sclerosis. Nature 2011;476:214–19.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Ascherio A,
      2. Munger KL
      . Environmental risk factors for multiple sclerosis. Part I: the role of infection. Ann  eurol 2007;61:288–99.
      OpenUrl
    5. ↵
      1. Ascherio A,
      2. Munger KL
      . Environmental risk factors for multiple sclerosis. Part II: noninfectious factors. Ann  eurol 2007;61:504–13.
      OpenUrl
    6. ↵
      1. Hedstrom AK,
      2. Olsson T,
      3. Alfredsson L
      . High body mass index before age 20 is associated with increased risk for multiple sclerosis in both men and women. Mult Scler 2012;18:1334–6.
      OpenUrlAbstract/FREE Full Text
    7. ↵
      1. Munger KL,
      2. Bentzen J,
      3. Laursen B,
      4. et al
      . Childhood body mass index and multiple sclerosis risk: a long-term cohort study. Mult Scler 2013;19:1323–9.
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Marmot M
      . Social determinants of health inequalities. Lancet 2005;365:1099–104.
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      1. Tabassum F,
      2. Kumari M,
      3. Rumley A,
      4. et al
      . Effects of socioeconomic position on inflammatory and hemostatic markers: a life-course analysis in the 1958 British birth cohort. Am J Epidemiol 2008;167:1332–41.
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. Loucks EB,
      2. Pilote L,
      3. Lynch JW,
      4. et al
      . Life course socioeconomic position is associated with inflammatory markers: the Framingham Offspring Study. Soc Sci Med 2010;71:187–95.
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Miller G,
      2. Chen E
      . Unfavorable socioeconomic conditions in early life presage expression of proinflammatory phenotype in adolescence. Psychosom Med 2007;69:402–9.
      OpenUrlAbstract/FREE Full Text
    12. ↵
      1. Miller GE,
      2. Chen E,
      3. Fok AK,
      4. et al
      . Low early-life social class leaves a biological residue manifested by decreased glucocorticoid and increased proinflammatory signaling. Proc Natl Acad Sci USA 2009;106:14716–21.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Antonovsky A,
      2. Leibowitz U,
      3. Medalie JM,
      4. et al
      . Reappraisal of possible etiologic factors in multiple sclerosis. Am J Public Health Nations Health 1968;58:836–48.
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Beebe GW,
      2. Kurtzke JF,
      3. Kurland LT,
      4. et al
      . Studies on the natural history of multiple sclerosis. 3. Epidemiologic analysis of the army experience in World War II. Neurology 1967;17:1–17.
      OpenUrl
    15. ↵
      1. Kurtzke JF,
      2. Page WF
      . Epidemiology of multiple sclerosis in US veterans: VII. Risk factors for MS. Neurology 1997;48:204–13.
      OpenUrlCrossRef
    16. ↵
      1. Ghadirian P,
      2. Dadgostar B,
      3. Azani R,
      4. et al
      . A case-control study of the association between socio-demographic, lifestyle and medical history factors and multiple sclerosis. Can J Public Health 2001;92:281–5.
      OpenUrlPubMedWeb of Science
    17. ↵
      1. Gardener H,
      2. Munger KL,
      3. Chitnis T,
      4. et al
      . Prenatal and perinatal factors and risk of multiple sclerosis. Epidemiology 2009;20:611–18.
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      1. Riise T,
      2. Kirkeleit J,
      3. Aarseth JH,
      4. et al
      . Risk of MS is not associated with exposure to crude oil, but increases with low level of education. Mult Scler 2011;17:780–7.
      OpenUrlAbstract/FREE Full Text
    19. ↵
      1. Nielsen NM,
      2. Jorgensen KT,
      3. Bager P,
      4. et al
      . Socioeconomic factors in childhood and the risk of multiple sclerosis. Am J Epidemiol 2013;177:1289–95.
      OpenUrlAbstract/FREE Full Text
    20. ↵
      1. Ponsonby AL,
      2. van der Mei I,
      3. Dwyer T,
      4. et al
      . Exposure to infant siblings during early life and risk of multiple sclerosis. Jama 2005;293:463–9.
      OpenUrlCrossRefPubMedWeb of Science
    21. ↵
      1. Islam T,
      2. Gauderman WJ,
      3. Cozen W,
      4. et al
      . Childhood sun exposure influences risk of multiple sclerosis in monozygotic twins. Neurology 2007;69:381–8.
      OpenUrlCrossRef
    22. ↵
      1. Goodin DS
      . The causal cascade to multiple sclerosis: a model for MS pathogenesis. PLoS ONE 2009;4:e4565.
      OpenUrlCrossRefPubMed
    23. ↵
      1. Merkin SS,
      2. Basurto-Davila R,
      3. Karlamangla A,
      4. et al
      . Neighborhoods and cumulative biological risk profiles by race/ethnicity in a national sample of U.S. adults: NHANES III. Ann Epidemiol 2009;19:194–201.
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Theall KP,
      2. Drury SS,
      3. Shirtcliff EA
      . Cumulative neighborhood risk of psychosocial stress and allostatic load in adolescents. Am J Epidemiol 2012;176(Suppl 7):S164–74.
      OpenUrlAbstract/FREE Full Text
    25. ↵
      1. Krieger N
      . Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health 1992;82:703–10.
      OpenUrlCrossRefPubMedWeb of Science
    26. ↵
      1. McDonald WI,
      2. Compston A,
      3. Edan G,
      4. et al
      . Recommended diagnostic criteria for multiple sclerosis: guidelines from the International Panel on the diagnosis of multiple sclerosis. Ann Neurol 2001;50:121–7.
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      1. Polman CH,
      2. Reingold SC,
      3. Edan G,
      4. et al
      . Diagnostic criteria for multiple sclerosis: 2005 revisions to the “McDonald Criteria”. Ann Neurol 2005;58:840–6.
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Kuh D,
      2. Ben-Shlomo Y,
      3. Lynch J,
      4. et al
      . Life course epidemiology. J Epidemiol Community Health 2003;57:778–83.
      OpenUrlAbstract/FREE Full Text
    29. ↵
      1. Tolppanen AM,
      2. Fraser A,
      3. Fraser WD,
      4. et al
      . Risk factors for variation in 25-hydroxyvitamin D(3) and D(2) concentrations and vitamin D deficiency in children. J Clin Endocrinol Metab 2012;97:1202–10.
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      1. De Jager PL,
      2. Chibnik LB,
      3. Cui J,
      4. et al
      . Integration of genetic risk factors into a clinical algorithm for multiple sclerosis susceptibility: a weighted genetic risk score. Lancet Neurol 2009;8:1111–19.
      OpenUrlCrossRefPubMedWeb of Science
    31. ↵
      1. Roden DM,
      2. Tyndale RF
      . Genomic medicine, precision medicine, personalized medicine: what's in a name? Clin Pharmacol Ther 2013;94:169–72.
      OpenUrlCrossRefPubMed
    32. ↵
      1. Gruenewald TL,
      2. Karlamangla AS,
      3. Hu P,
      4. et al
      . History of socioeconomic disadvantage and allostatic load in later life. Soc Sci Med 2012;74:75–83.
      OpenUrlCrossRefPubMed
    33. ↵
      1. Krieger N,
      2. Williams DR,
      3. Moss NE
      . Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health 1997;18:341–78.
      OpenUrlCrossRefPubMedWeb of Science
    34. ↵
      1. Galobardes B,
      2. Shaw M,
      3. Lawlor DA,
      4. et al
      . Indicators of socioeconomic position (part 2). J Epidemiol Community Health 2006;60:95–101.
      OpenUrlAbstract/FREE Full Text
    35. ↵
      1. Galobardes B,
      2. Shaw M,
      3. Lawlor DA,
      4. et al
      . Indicators of socioeconomic position (part 1). J Epidemiol Community Health 2006;60:7–12.
      OpenUrlAbstract/FREE Full Text
    36. ↵
      1. Hauser RM
      . Measuring socioeconomic status in studies of child development. Child development 1994;65:1541–5.
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      1. Long JE,
      2. Caudill SB
      . Racial differences in homeownership and housing wealth. Econ Inquiry 1992;30:83–100.
      OpenUrlCrossRefWeb of Science
    38. ↵
      1. Poulton R,
      2. Caspi A,
      3. Milne BJ,
      4. et al
      . Association between children's experience of socioeconomic disadvantage and adult health: a life-course study. Lancet 2002;360:1640–5.
      OpenUrlCrossRefPubMedWeb of Science
    39. ↵
      1. Viner RM,
      2. Ozer EM,
      3. Denny S,
      4. et al
      . Adolescence and the social determinants of health. Lancet 2012;379:1641–52.
      OpenUrlCrossRefPubMedWeb of Science
    40. ↵
      1. Lynch JW,
      2. Kaplan GA,
      3. Salonen JT
      . Why do poor people behave poorly? Variation in adult health behaviours and psychosocial characteristics by stages of the socioeconomic lifecourse. Soc Sci Med 1997;44:809–19.
      OpenUrlCrossRefPubMedWeb of Science
    41. ↵
      1. Mishra G,
      2. Nitsch D,
      3. Black S,
      4. et al
      . A structured approach to modelling the effects of binary exposure variables over the life course. Int J Epidemiol 2009;38:528–37.
      OpenUrlAbstract/FREE Full Text
    42. ↵
      1. Grzywacz JG,
      2. Almeida DM,
      3. Neupert SD,
      4. et al
      . Socioeconomic status and health: a micro-level analysis of exposure and vulnerability to daily stressors. J Health Soc Behav 2004;45:1–16.
      OpenUrlAbstract/FREE Full Text
    View Abstract

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • CS and LFB are contributed equal.

    • Contributors All authors made substantial contributions to (1) conception and design (FBSB, LFB and CS) or acquisition of data (LFB, BSA, LS, KHB, PPR, HQ, AB and CS) or analysis and interpretation of data (FBSB, LFB and CS); (2) drafting the article (FBSB and LFB) or reviewing it and, if appropriate, revising it critically for important intellectual content (BSA, LS, PPR, HQ, AB and CS); and (3) final approval of the version to be published (all authors). All authors participated sufficiently in the work to believe in its overall validity and take public responsibility for appropriate portions of its content.

    • Funding NIH—the development of the KPNC MS Register; NMSS—postdoctoral fellowship.

    • Competing interests None.

    • Patient consent Obtained.

    • Ethics approval Institutional Review Boards of Kaiser Permanente, Division of Research, and the University of California, Berkeley.

    • Provenance and peer review Not commissioned; externally peer reviewed.