SMOKING and MULTIPLE SCLEROSIS: a SYSTEMATIC REVIEW and META-ANALYSIS USING the BRADFORD HILL CRITERIA for CAUSATION a Thesis

Home , Bradford Hill criteria

SMOKING AND MULTIPLE SCLEROSIS: A SYSTEMATIC REVIEW AND

META-ANALYSIS USING THE BRADFORD HILL CRITERIA FOR CAUSATION

A Thesis

Submitted to the Faculty of Graduate Studies and Research

In Partial Fulfillment of the Requirements

For the Degree of

Master of Science

Kinesiology and Health Studies

University of Regina

Michelle Leigh Degelman

Regina, Saskatchewan

March, 2017

UNIVERSITY OF REGINA

FACULTY OF GRADUATE STUDIES AND RESEARCH

SUPERVISORY AND EXAMINING COMMITTEE

Michelle Leigh Degelman, candidate for the degree of Master of Science in Kinesiology and Health Studies, has presented a thesis titled, Smoking and Multiple Sclerosis: A Systematic Review and Meta-Analysis Using the Bradford Hill Criteria for Causation, in an oral examination held on December 14, 2016. The following committee members have found the thesis acceptable in form and content, and that the candidate demonstrated satisfactory knowledge of the subject material.

External Examiner: *Dr. Ruth Ann Marrie, University of Manitoba

Supervisor: Dr. Katya Herman, Faculty of Kinesiology and Health Studies

Committee Member: Dr. Kim Dorsch, Faculty of Kinesiology and Health Studies

Committee Member: Dr. Donald Sharpe, Department of Psychology

Chair of Defense: Dr. Elizabeth Domm, Faculty of Nursing

*Via teleconference

Abstract

Background. Multiple sclerosis (MS) is one of the most common neurological disorders in the world. However, the cause(s) of MS is unknown. Studies have suggested that smoking may be related to the development and progression of MS. The Bradford Hill criteria are commonly used to assess potentially causal associations. However, to date, no systematic reviews and meta-analyses had used these criteria to comprehensively evaluate the relationship between smoking and MS. The objective of this research was to assess the relationship between smoking and both MS risk and MS progression through a systematic review and meta-analysis, subsequently applying Hill’s criteria to further assess the likelihood of these associations being causal. Methods. Databases including

Medline, EMBASE, CINAHL, PsycInfo, and Cochrane Library were searched for relevant studies up until July 28, 2015. A random-effects meta-analysis was conducted for three outcomes: MS risk, conversion from clinically isolated syndrome (CIS) to clinically definite multiple sclerosis (CDMS), and progression from relapsing-remitting multiple sclerosis (RRMS) to secondary-progressive multiple sclerosis (SPMS).

Consideration was given to dose-response relationships and risk factor interactions, and discussions of mechanisms and analogous associations presented in the studies. Hill’s criteria were applied to assess causality of the relationships between smoking and each outcome. The effect of second-hand smoke exposure was also briefly reviewed. Results.

Smoking had a statistically significant association with both MS risk (conservative:

OR/RR 1.54, 95% CI [1.46-1.63]) and SPMS risk (HR 1.80, 95% CI [1.04-3.10]), but the association with progression from CIS to CDMS was non-significant (HR 1.13, 95% CI

[0.73-1.76]). Based on Hill’s criteria, there was strong evidence of a causal role of

i smoking in MS risk, but only moderate evidence of a causal association between smoking and MS progression. Heterogeneity in study designs and target populations, inconsistent results, and an overall scarcity of studies point to the need for more research in the area of second-hand smoke exposure in relation to MS prior to conducting a detailed meta- analysis. Conclusion. This first review to supplement systematic review and meta- analytic methods with the application of Hill’s criteria to analyze the smoking-MS association provided evidence supporting the causal involvement of smoking in the development and progression of MS. As such, smoking cessation programs and policies should highlight MS as an additional health risk to deter the public from smoking.

Keywords: Multiple sclerosis; cigarette smoking; clinically isolated syndrome; Bradford

Hill

ii Acknowledgements

First, I would like to thank my supervisor, Dr. Katya Herman (Kinesiology and

Health Studies) for the countless hours she devoted to helping me. Over the past couple of years, my writing and presentation skills have vastly improved, as have my abilities as a researcher and collaborator. Thank you Dr. Herman for encouraging me to accept challenging opportunities, from presenting my own research at a national conference to being interviewed by the Leader Post and CBC Radio. These experiences will benefit me as I embark on a career in the health research field. In short, any graduate student would be truly fortunate to have Dr. Herman as their supervisor.

I would also like to thank my committee members, Dr. Kim Dorsch (Kinesiology and Health Studies) and Dr. Donald Sharpe (Psychology). Dr. Dorsch, thank you for the invaluable guidance and support over the course of my Masters program, as well as all the time committed to helping me with undergraduate and graduate research projects. I truly enjoyed our time collaborating, and welcome the opportunity to work together in the near future. Dr. Sharpe, thank you for familiarizing me with the meta-analysis process, as well as being readily available to answer questions. The impressive time commitment and constructive feedback were instrumental in the successful completion of this project.

I am thankful for the financial support received from the Canadian Institutes of

Health Research in the form of a Frederick Banting and Charles Best Canada Graduate

Scholarship.

iii Table of Contents

Abstract ...... i

Acknowledgements ...... iii

Table of Contents ...... iv

List of Tables ...... vi

List of Figures ...... vii

List of Abbreviations ...... viii

CHAPTER 1: General Introduction ...... 1 1.1 Introduction ...... 1 1.1.1 Definitions and Concepts ...... 2 1.1.2 What is Known – What is Needed ...... 4

1.1.3 Conceptual Framework ...... 8 1.2 Overview and Overall Objective ...... 10 1.2.1 Specific Objectives ...... 10 1.3 References ...... 12

CHAPTER 2: Review of the Literature ...... 18 2.1 Multiple Sclerosis ...... 18 2.1.1 Symptoms ...... 18 2.1.2 Diagnosis...... 20 2.1.3 Types ...... 22 2.1.4 Impact ...... 23 2.1.5 Etiology ...... 25 2.1.5.1 Smoking ...... 26 2.2 Overview of Systematic Review and Meta-Analytic Methods ...... 31 2.3 Previous Use of Hill’s Criteria ...... 35 2.3.1 Previous Use of Hill’s Criteria in Narrative Reviews of MS Causation ...... 35 2.3.2 Previous Use of Hill’s Criteria in Systematic Reviews and Meta-Analyses ...... 38 2.4 Summary ...... 39 2.5 References ...... 41

CHAPTER 3: Manuscript - Smoking and Multiple Sclerosis: A Systematic Review and Meta-

Analysis Using the Bradford Hill Criteria for Causation...... 54 Abstract ...... 55 Introduction ...... 57 Methods...... 59 Results ...... 67 Discussion ...... 83 References ...... 93

CHAPTER 4: General Summary and Future Directions ...... 104 4.1 References ...... 108

APPENDICES ...... 109 APPENDIX A: Title, Abstract, & Full-Text Screening Questions Specific to Thesis ....110 APPENDIX B: Studies Included in Smoking-MS Risk Meta-Analysis and Corresponding Effect Estimates ...... 111 APPENDIX C: GRADE Approach ...... 113 APPENDIX D: Included Studies in Each Quantitative and Qualitative Analysis ...... 114 APPENDIX E: Details of Studies Included in Meta-Analysis ...... 123 APPENDIX F: Excluded Full-Text Studies, with Corresponding Reasons ...... 127

List of Tables

CHAPTER 3: Manuscript

Table 1. Smoking-MS risk effect estimates with 95% CIs for study subgroups ...... 75

Table 2. Quality of evidence evaluation for MS risk and MS progression outcomes using the GRADE approach ...... 76

Table 3. Causality evaluation of the association of smoking with MS risk and MS progression outcomes using Hill’s criteria ...... 82

List of Figures

CHAPTER 1: General Introduction

Figure 1-1. Conceptual scheme for the causes of a hypothetical disease ...... 9

CHAPTER 3: Manuscript

Figure 1. Flow diagram of study selection process depicting number of studies identified, excluded (including titles, abstracts, full-text articles), and included in review ...... 68

Figure 2. Meta-analysis and forest plot: Relationship between smoking and MS risk (conservative) ...... 70

Figure 3. Meta-analysis and forest plot: Relationship between smoking and MS risk (non-conservative) ...... 71

Figure 4. Funnel plot: Conservative analysis, association of smoking with MS risk ...... 72

Figure 5. Funnel plot: Non-conservative analysis, association of smoking with MS risk ..... 73

Figure 6. Meta-analysis and forest plot: Relationship between smoking and conversion from CIS to CDMS ...... 78

Figure 7. Meta-analysis and forest plot: Relationship between smoking and conversion from RRMS to SPMS ...... 80

CHAPTER 4: General Summary and Future Directions

Figure 4-1. Potential causal pathways for MS using currently suspected component causes...... 105

vii

List of Abbreviations

CCSVI Chronic Cerebrospinal Venous Insufficiency

CDMS Clinically Definite Multiple Sclerosis

CI Confidence Interval

CIS Clinically Isolated Syndrome

CNS Central Nervous System

EBV Epstein-Barr Virus

GRADE Grading of Recommendations Assessment, Development, and Evaluation

HLA Human Leukocyte Antigen

HR Hazard Ratio

IV Inverse Variance

MS Multiple Sclerosis

NARCOMS North American Research Committee on Multiple Sclerosis

OR Odds Ratio

PPMS Primary-Progressive Multiple Sclerosis

PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses

RR Risk Ratio

RRMS Relapsing-Remitting Multiple Sclerosis

SE Standard Error

SPMS Secondary-Progressive Multiple Sclerosis

viii

CHAPTER 1: General Introduction

1.1 Introduction

Multiple sclerosis (MS) is a common neurological disorder of the central nervous system (CNS).1 The prevalence of MS in Canada is one of the highest in the world at 291 cases per 100,000 people compared to the global median prevalence of 33 cases per

100,000 people.2 MS is an extremely unpredictable disease with a challenging diagnosis as it exhibits many different symptoms and a variable disease course.3 Furthermore, the disease has been shown to have a profound adverse effect on the well-being of patients, family members, and society in general.4-6

The cause of MS remains unknown.7 However, a number of factors, including genetics, infectious agents, and lifestyle behaviours, have been shown to increase one’s risk of developing MS.8 Included among lifestyle behaviours is smoking, a risk factor that has been increasingly studied over the past decade.9-11

The Bradford Hill criteria for causation12 encompass a list of nine criteria that are widely known for evaluating causation between an exposure and a given outcome. Five previous systematic reviews and meta-analyses,13-17 as well as two systematic reviews without meta-analyses,18, 19 have been published that examine the relationship between smoking and MS. However, none of the reviews methodically used Hill’s criteria to evaluate the relationship between smoking and MS. While two recent narrative reviews used Hill’s criteria to examine this relationship,20, 21 they lacked systematic methods.

Conducting an updated systematic review and meta-analysis and subsequently using

Hill’s criteria to examine the association between smoking and MS would provide

1 valuable insight into this relationship as this combination of methods makes it possible to evaluate causation at a more comprehensive level.

1.1.1 Definitions and Concepts

MS is an autoimmune disorder of the CNS.1 The disease process is primarily characterized by peripheral immune activation whereby activated immune cells in the blood cross the blood-brain barrier, triggering an immune reaction that leads to patchy myelin damage and axonal injury.1 The demyelination, inflammation, and scarring are what characterize lesions (or plaques).1 Myelin, the tissue that wraps around the nerve axons in the CNS, is responsible for conducting impulses quickly along nerves to different parts of the body, but lesions disrupt this conduction which ultimately leads to the symptoms of MS.1

A risk factor, a term first coined by the original investigators of the Framingham

Heart Study,22 is defined as an “exposure, behaviour, or attribute that, if present and active, clearly increases the probability of a particular disease in a group of people who have the risk factor compared with an otherwise similar group of people who do not”.23 A cause of a particular disease has been defined as an “antecedent event, condition, or characteristic that was necessary for the occurrence of the disease at the moment it occurred, given that other conditions are fixed”.24

Smoking is generally described as “the inhalation of the smoke of burning tobacco of cigarettes, pipes or cigars, [which] may be an occasional habit or, more often, a smoking habit involving a physical addiction to tobacco products, primarily nicotine”.25

Smoking is considered the most important avoidable health risk in developed countries,

2 and is linked to a variety of diseases such as cancer, coronary heart disease, stroke, chronic obstructive pulmonary disease, and peptic ulcer disease, as well as abnormal fetal and neonatal growth and development during pregnancy.26 Furthermore, second-hand smoke or environmental tobacco smoke, which involves the inhalation of the smoke of burning tobacco by non-smokers, is linked to many of the same smoking-related diseases as previously stated.27

Systematic reviews strive to “collate all empirical evidence that fits pre-specified eligibility criteria in order to answer a specific research question. [These reviews use] explicit, systematic methods that are selected with a view to minimizing bias, thus providing more reliable findings from which conclusions can be drawn and decisions made”.28 In general, a systematic review involves locating and assessing all relevant studies, and obtaining and synthesizing data from such studies. A systematic review may or may not incorporate a meta-analysis, which involves “the use of statistical methods to

[combine] the results of independent studies [and therefore] … can provide more precise

[effect] estimates … than those derived from the individual studies included within a review”.28 One meta-analysis does not necessarily need to combine the results of all of the studies included in a systematic review, especially if studies are at risk of bias or are varied in terms of exposures and outcomes.28 Such a meta-analysis would likely generate an erroneous effect estimate. Given that systematic reviews consider the entire research literature available on a selected topic, and meta-analyses summarize the quantitative results from such studies, systematic reviews and meta-analyses are important for establishing clinical practice guidelines and keeping health care providers informed of up-to-date medical evidence.29, 30

3 1.1.2 What is Known – What is Needed

A potential MS risk factor that is currently receiving considerable attention is smoking. At the population-level, it has been demonstrated that smoking prevalence is associated with MS prevalence to some degree.31 In addition, several studies have shown that smoking is generally associated with increased MS risk and severe and rapid disease progression, with greater risk and progression associated with greater smoking exposure.32-35 Conversely, smoking cessation is generally associated with a lower risk of developing MS as well as a slower disease progression.33, 36 The following are some progression outcomes that have been associated with smoking: 1) Conversion from a first episode of CNS demyelination, known as clinically isolated syndrome (CIS), to clinically definite multiple sclerosis (CDMS);37 2) disease progression as measured by the conversion from a relapsing-remitting multiple sclerosis (RRMS) disease course to a secondary-progressive multiple sclerosis (SPMS) disease course;32 3) progression of disability;36 and 4) progression of disease activity, including lesions.32

Three previous systematic reviews and meta-analyses published in 2007,14 2011,13 and 201415 concluded that smoking was associated with MS risk. The review conducted in 2011 involved a further analysis examining MS progression and showed a statistically non-significant relationship between smoking and SPMS risk.13 Two more recent systematic reviews with meta-analyses published in 2016 concluded that different smoking statuses (current, past, and ever-smoking, the latter comprised of current and past smokers in combination) were associated with a higher risk of MS.16, 17 While one of the reviews showed a relationship between passive smoke exposure and MS risk,17 the other review did not.16 In addition, two systematic reviews without meta-analyses were

4 recently published, one of which showed a relationship between smoking and MS risk and progression,18 while the other review, exploring MS risk in more detail, did not demonstrate a consistent relationship between smoking and a certain course of MS.19

There are also several narrative reviews examining the association between smoking and

MS,20, 21, 38, 39 all of which provided evidence of the adverse effects of smoking on MS risk and/or progression. While there is an abundance of research on the smoking-MS relationship, there is a critical need to comprehensively evaluate the likelihood that smoking is truly playing a causal role in MS.

The Bradford Hill criteria for causation12 are used to evaluate a causal association between an exposure and a particular disease. Hill listed nine criteria that together increase the probability that an association is causal. The criteria are described below:

1. Strength: stronger relationships between an exposure and a disease have a greater

likelihood of being causal compared to weaker relationships; strength is commonly

conveyed by an odds ratio (OR) or risk ratio (RR), whereby a larger magnitude of the

estimate indicates a greater probability of the association being causal, though

weaker associations do not necessarily indicate a non-causal relationship.

2. Consistency: repetition of a relationship across studies undertaken by different

researchers, in different places and times, and using different population groups and

methods, strengthens the likelihood that the relationship is causal.

3. Specificity: a causal association is more likely to exist if an exposure is limited to

one disease; however, the importance of this criterion should not be over-

emphasized, given that one-to-one relationships are rare and generally non-

applicable to most chronic diseases.

5 4. Temporality: an exposure must always precede the disease in time in order for a

causal association to exist; this is the one criterion that always must be met in order

to infer causation.

5. Biological Gradient: the presence of a dose-response curve strengthens the

likelihood of a causal relationship (i.e., an increase in the level of an exposure is

followed by an increase in disease risk).

6. Plausibility: a causal relationship is consistent with current biological knowledge of

disease processes (i.e., it is supported by known biological mechanisms underlying

the relationship); however, plausibility is limited by existing biological knowledge,

and as such, the absence of knowledge to support a new relationship does not

necessarily indicate a non-causal relationship.

7. Coherence: a causal relationship will likely not deviate from known facts related to

the natural history and biology of the disease.

8. Experimental Evidence: a causal relationship is more likely to exist if experimental

evidence under controlled conditions shows that changing the exposure results in a

change in the outcome. In many cases, it is not possible to obtain experimental

evidence ethically when examining disease causation.

9. Analogy: a causal relationship is more likely to be accepted if elsewhere a similar

exposure results in a similar outcome.

Hill developed these criteria during the rise of non-communicable disease epidemiology.40 A portion of these criteria was used in assessing the relationship between cigarette smoking and lung cancer in a landmark report to the Surgeon General of the

United States.41 Hill’s criteria are still viewed as a valuable tool against which

6 characteristics of an association can be tested to determine if causal inference is a reasonable conclusion.40 The criteria have been applied in many areas of epidemiology to establish support for causality, such as the link between Vitamin D and breast cancer,42 dietary acid and osteoporosis,43 and alcohol consumption and cardiovascular disease outcomes.44 In addition, Hill’s criteria have been used in narrative reviews to independently assess the causality of Vitamin D insufficiency,45 the Hepatitis B vaccine,46 and smoking20, 21 in relation to MS. Overall, relationships between an exposure and a particular disease that meet a greater number of criteria have a greater likelihood of being causal than those meeting fewer criteria.47

However, the Bradford Hill criteria have attracted some criticism. Hill himself noted that the criteria do not provide absolute evidence for determining whether a relationship is causal.12 For instance, associations that do not satisfy the criteria do not necessarily provide definite support for non-causation (e.g., causal relationships can be weak, non-specific, inconsistent, odd). Similarly, associations satisfying Hill’s criteria do not provide definite support for causation (i.e., a strong relationship could be due to a strong underlying confounder). Building on Hill’s argument, there is agreement that satisfying the criteria neither deductively nor inductively supports a causal claim.48

Specifically, one cannot detect the aspects of the criteria that guarantee that a causal claim is true (deduction) or that a false claim is improbable (induction). Furthermore, given that Hill’s criteria are used to analyze epidemiological associations that may be the result of bias or confounding, such data are not likely to undeniably demonstrate causation.49 It has also been argued that the application of Hill’s criteria involves a complex causal system with many components rather than a simple one-to-one

7 relationship, limiting the value of the criteria.50 Nevertheless, the Bradford Hill criteria are important in establishing justified causal links between an exposure and a given disease that represents the best explanation based on study data.48, 49

To date, two narrative reviews used Hill’s criteria in their reviews of studies published up to and including 201121 and 201420 respectively. Arguably, utilizing Hill’s criteria to extensively evaluate the causal nature of the smoking-MS relationship based on the most reliable findings identified in a rigorous systematic review and meta-analysis provides even greater insight into this relationship, and would greatly assist in establishing if smoking is playing a causal role in MS.

1.1.3 Conceptual Framework

The conceptual framework used as the underlying basis for this review of smoking as a cause for MS is Rothman’s Model of Necessary and Sufficient Causes,51 a well-known model conceptualizing chronic disease etiology presented in Figure 1-1.

Rothman’s model shows how a hypothetical disease may have multiple sufficient causes (I, II, III, etc.), each consisting of different combinations of component causes (A,

B, C, etc.). Any component cause that is removed from a sufficient cause makes the remaining components insufficient, thus preventing disease onset by that sufficient cause.

Furthermore, a necessary cause is a component cause that appears in every sufficient cause (e.g., A in Figure 1-1). As such, researchers always hope to identify a necessary cause of the disease in question as removing it would lead to complete disease prevention.

Figure 1-1. Conceptual scheme for the causes of a hypothetical disease. Reprinted from “Causes,” by K. J. Rothman, 1976, Am J Epidemiol, 104(6), p. 589.

9 MS fits well into Rothman’s model given that this chronic disease is considered to be a multi-causal process. In other words, there is no single factor that has been found to be sufficient to cause MS.2 However, a necessary cause of MS has yet to be identified.7

With regards to smoking, it is neither a necessary cause of MS (i.e., smoking is not represented by A in Figure 1-1 since non-smokers can develop MS) nor is smoking a sufficient cause of MS (i.e., I, II, III in Figure 1-1 do not consist solely of smoking since not all smokers develop MS). Therefore, it is envisioned that smoking could represent a component cause that may be part of one or several sufficient causes of MS (i.e., smoking could be represented by B, C, D, etc. in Figure 1-1).

1.2 Overview and Overall Objective

Hill’s criteria have previously been used in two narrative reviews to evaluate the association between smoking and MS.20, 21 However, the criteria have not been used in an updated systematic review and meta-analysis to evaluate this relationship. The overall objective of this research was to examine the relationship between smoking and MS through a systematic review and meta-analysis, incorporating the Bradford Hill criteria for causation.

1.2.1 Specific Objectives

1) To analyze the relationship between smoking and MS risk in a systematic review

and meta-analysis that incorporates the Bradford Hill criteria for causation;

2) To analyze the relationship between smoking and MS progression, including the

conversion from CIS to CDMS, and the conversion from RRMS to SPMS, in a

10 systematic review and meta-analysis that incorporates the Bradford Hill criteria

for causation; and

3) To provide a brief summary of the literature as it pertains to the relationship

between other forms of smoke exposure (i.e., passive, prenatal) and MS, in the

form of a systematic review.

Rationale: The present research serves to expand on previous systematic reviews and meta-analyses that have examined the effect of smoking on MS risk. Additional relevant studies are included and a more detailed evaluation of causation is conducted.

There is also a need to examine the relationship between smoking and MS progression in greater detail, given that only one previous systematic review and meta- analysis analyzed the effect of smoking on the conversion from RRMS to SPMS,13 and none examined the influence of smoking on the conversion from CIS to CDMS. This comprehensive examination will help clarify the role of smoking in important stages of

MS. To date, Hill’s criteria have not been methodically used in a systematic review and meta-analysis to evaluate the association between smoking and MS. In the present review, Hill’s criteria provide a basis for thoroughly evaluating the causal nature of the relationship, offering new insight into the role of smoking in MS.

Two recent reviews were the first to use both systematic review and meta-analytic methods to evaluate the relationship between passive smoking and MS.16, 17 Given this new research on second-hand smoke exposure and the possibility that such exposures are important in the development of MS, providing a summary of the literature to date in this area, as well as summarizing the MS literature on other forms of smoke exposure (e.g., prenatal smoking exposure) is essential for stimulating further research.

11 1.3 References

1. Murray TJ. Multiple sclerosis: The history of a disease. New York, NY: Demos

Medical Publishing, 2005.

2. Multiple Sclerosis International Federation. Atlas of MS 2013. London, UK: Multiple

Sclerosis International Federation, 2013.

3. Hurwitz BJ. The diagnosis of multiple sclerosis and the clinical subtypes. Ann Indian

Acad Neurol 2009;12(4):226-30.

4. Buchanan RJ, Radin D, Huang C. Caregiver burden among informal caregivers assisting people with multiple sclerosis. Int J MS Care 2011;13(2):76-83.

5. Isaksson AK, Ahlstrom G. From symptom to diagnosis: Illness experiences of multiple sclerosis patients. J Neurosci Nurs 2006;38(4):229-37.

6. Karampampa K, Gustavsson A, Miltenburger C, Kindundu CM, Selchen DH.

Treatment experience, burden, and unmet needs (TRIBUNE) in multiple sclerosis: The costs and utilities of MS patients in Canada. J Popul Ther Clin Pharmacol

2012;19(1):e11-25.

7. World Health Organization. Atlas: Multiple sclerosis resources in the world. Geneva,

CH: WHO Press, 2008.

8. Talley CL. A history of multiple sclerosis. Westport, CT: Praeger, 2008.

9. Hernan MA, Jick SS, Logroscino G, Olek MJ, Ascherio A, Jick H. Cigarette smoking and the progression of multiple sclerosis. Brain 2005;128(Pt 6):1461-5.

10. O'Gorman C, Bukhari W, Todd A, Freeman S, Broadley SA. Smoking increases the risk of multiple sclerosis in Queensland, Australia. J Clin Neurosci 2014;21(10):1730-3.

12 11. Salzer J, Hallmans G, Nystrom M, Stenlund H, Wadell G, Sundstrom P. Smoking as a risk factor for multiple sclerosis. Mult Scler 2013;19(8):1022-7.

12. Hill AB. The environment and disease: Association or causation? Proc R Soc Med

1965;58:295-300.

13. Handel AE, Williamson AJ, Disanto G, Dobson R, Giovannoni G, Ramagopalan SV.

Smoking and multiple sclerosis: An updated meta-analysis. PLoS One 2011;6(1):e16149.

14. Hawkes CH. Smoking is a risk factor for multiple sclerosis: A metanalysis. Mult

Scler 2007;13(5):610-5.

15. O'Gorman C, Broadley SA. Smoking and multiple sclerosis: Evidence for latitudinal and temporal variation. J Neurol 2014;261(9):1677-83.

16. Poorolajal J, Bahrami M, Karami M, Hooshmand E. Effect of smoking on multiple sclerosis: A meta-analysis. J Public Health (Oxf) 2016. doi:10.1093/pubmed/fdw030.

17. Zhang P, Wang R, Li Z, et al. The risk of smoking on multiple sclerosis: A meta- analysis based on 20,626 cases from case-control and cohort studies. PeerJ

2016;4:e1797.

18. McKay KA, Jahanfar S, Duggan T, Tkachuk S, Tremlett H. Factors associated with onset, relapses or progression in multiple sclerosis: A systematic review. Neurotoxicology

2016 (in press).

19. McKay KA, Kwan V, Duggan T, Tremlett H. Risk factors associated with the onset of relapsing-remitting and primary progressive multiple sclerosis: A systematic review.

Biomed Res Int 2015;2015:817238.

13 20. Loken-Amsrud KI, Lossius A, Torkildsen O, Holmoy T. Impact of the environment on multiple sclerosis. Tidsskr Nor Laegeforen 2015;135(9):856-60.

21. Wingerchuk DM. Smoking: Effects on multiple sclerosis susceptibility and disease progression. Ther Adv Neurol Disord 2012;5(1):13-22.

22. Kannel WB, Dawber TR, Kagan A, Revotskie N, Stokes J, 3rd. Factors of risk in the development of coronary heart disease--six year follow-up experience. The Framingham

Study. Ann Intern Med 1961;55:33-50.

23. Jekel JF, Katz DL, Wild D, Elmore JG. Epidemiology, biostatistics, and preventative medicine. Philadelphia, PA: Elsevier Health Sciences, 2007.

24. Rothman KJ, Greenland S. Causation and causal inference. In: Modern epidemiology.

2nd ed. Philadelphia, PA: Lippincott, 1998:7-28.

25. Leone A, Landini L. What is tobacco smoke? Sociocultural dimensions of the association with cardiovascular risk. Curr Pharm Des 2010;16(23):2510-7.

26. Fagerstrom K. The epidemiology of smoking: Health consequences and benefits of cessation. Drugs 2002;62 Suppl 2:1-9.

27. U.S. Department of Health and Human Services. The health consequences of involuntary exposure to tobacco smoke: A report of the Surgeon General. Atlanta, GA:

Centers for Disease Control and Prevention, 2006.

28. Higgins JPT, Green S (editors). Cochrane handbook for systematic reviews of interventions (version 5.1.0) [updated March 2011]. Cochrane Collaboration, 2011.

Available at: www.handbook.cochrane.org.

29. Akobeng AK. Understanding systematic reviews and meta-analysis. Arch Dis Child

2005;90(8):845-8.

14 30. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. PLoS Med 2009;6(7):e1000097.

31. Handel AE, Handunnetthi L, Giovannoni G, Ebers GC, Ramagopalan SV. Genetic and environmental factors and the distribution of multiple sclerosis in Europe. Eur J

Neurol 2010;17(9):1210-4.

32. Healy BC, Ali EN, Guttmann CR, et al. Smoking and disease progression in multiple sclerosis. Arch Neurol 2009;66(7):858-64.

33. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Smoking and multiple sclerosis susceptibility. Eur J Epidemiol 2013;28(11):867-74.

34. Hernan MA, Olek MJ, Ascherio A. Cigarette smoking and incidence of multiple sclerosis. Am J Epidemiol 2001;154(1):69-74.

35. Sundstrom P, Nystrom L. Smoking worsens the prognosis in multiple sclerosis. Mult

Scler 2008;14(8):1031-5.

36. Manouchehrinia A, Tench CR, Maxted J, Bibani RH, Britton J, Constantinescu CS.

Tobacco smoking and disability progression in multiple sclerosis: United Kingdom cohort study. Brain 2013;136(Pt 7):2298-304.

37. Di Pauli F, Reindl M, Ehling R, et al. Smoking is a risk factor for early conversion to clinically definite multiple sclerosis. Mult Scler 2008;14(8):1026-30.

38. Jafari N, Hintzen RQ. The association between cigarette smoking and multiple sclerosis. J Neurol Sci 2011;311(1-2):78-85.

39. Shirani A, Tremlett H. The effect of smoking on the symptoms and progression of multiple sclerosis: A review. J Inflamm Res 2010;3:115-26.

15 40. Lucas RM, McMichael AJ. Association or causation: Evaluating links between

"environment and disease". Bull World Health Organ 2005;83(10):792-5.

41. U.S. Department of Health, Education, and Welfare. Smoking and health: Report of the Advisory Committee to the Surgeon General of the Public Health Service.

Washington, D. C.: Government Printing Office, 1964.

42. Mohr SB, Gorham ED, Alcaraz JE, et al. Does the evidence for an inverse relationship between serum vitamin D status and breast cancer risk satisfy the Hill criteria? Dermatoendocrinol 2012;4(2):152-7.

43. Fenton TR, Tough SC, Lyon AW, Eliasziw M, Hanley DA. Causal assessment of dietary acid load and bone disease: A systematic review & meta-analysis applying Hill's epidemiologic criteria for causality. Nutr J 2011;10:41.

44. Ronksley PE, Brien SE, Turner BJ, Mukamal KJ, Ghali WA. Association of alcohol consumption with selected cardiovascular disease outcomes: A systematic review and meta-analysis. BMJ 2011;342:d671.

45. Hanwell HE, Banwell B. Assessment of evidence for a protective role of vitamin D in multiple sclerosis. Biochim Biophys Acta 2011;1812(2):202-12.

46. Le Houezec D. Evolution of multiple sclerosis in France since the beginning of hepatitis B vaccination. Immunol Res 2014;60(2-3):219-25.

47. Morabia A. On the origin of Hill's causal criteria. Epidemiology 1991;2(5):367-9.

48. Ward AC. The role of causal criteria in causal inferences: Bradford Hill's "aspects of association". Epidemiol Perspect Innov 2009;6:2.

49. Wakeford R. Association and causation in epidemiology - half a century since the publication of Bradford Hill's interpretational guidance. J R Soc Med 2015;108(1):4-6.

16 50. Hofler M. The Bradford Hill considerations on causality: A counterfactual perspective. Emerg Themes Epidemiol 2005;2:11.

51. Rothman KJ. Causes. Am J Epidemiol 1976;104(6):587-92.

17 CHAPTER 2: Review of the Literature

2.1 Multiple Sclerosis

MS is among the most common neurological diseases in the world, and one of the primary causes of non-traumatic disability in the young adult population.1 According to an extensive international study, approximately 2.3 million people were diagnosed with

MS worldwide in 2013.1 While MS is considered a global disease, numerous studies have confirmed that Canada has one of the highest prevalences of MS in the world,2-4 with some of the highest prevalences being reported in Alberta,5 Saskatchewan,6 Manitoba,7 and Nova Scotia.8 In 2013, over 97,000 Canadians were living with MS.1

Although MS affects every demographic, this chronic disease primarily has its onset in adults between 20 and 40 years of age,9 and is twice as common in women as it is in men.1 In addition, MS is more common among Caucasians with a northern European background than among Asians, Africans, and Native Americans.10 Studies have suggested the existence of a latitudinal gradient in MS prevalence, with countries located further away from the equator, such as Denmark, Ireland, and Canada, having higher MS prevalences compared to countries located closer to the equator, such as China, Thailand, and Argentina.4, 11 There is currently no cure for MS.12

2.1.1 Symptoms

MS is highly unpredictable as the disease has a variety of different symptoms, some of the most common being extreme fatigue and sensory disturbances.13 According to a study that used the North American Research Committee on Multiple Sclerosis

(NARCOMS) Registry, a database containing symptom information on 36,000 MS

18 patients, sensory symptoms (85%) and unexplained fatigue (81%) were the most common symptoms reported within the first year after disease onset.14 When it comes to sensory symptoms, many MS patients experience numbness, tingling, pricking, and sensitivity loss in the trunk as well as in the upper and lower extremities.15

Many MS patients will also experience motor impairment and vision problems.

Specifically, motor symptoms can occur in different parts of the body, including spasticity in the upper and lower limbs, and tremors occurring in the arms, legs, head, and trunk.16, 17 Such symptoms may lead to mobility issues, balance problems, postural instability, and wheelchair dependency.16, 18 NARCOMS participants reported some degree of spasticity (54%) and tremor (38%) within the first year of MS onset.14 In addition, 35% of participants reported that their mobility was noticeably affected in the first year, with an additional 15% of participants occasionally requiring a mobility device.14 With respect to vision problems, an eye condition known as optic neuritis is a frequent symptom of MS. Optic neuritis is an inflammation of one or both optic nerves, which usually leads to temporary loss of vision.19 Other visual disturbances that MS patients may experience include diplopia (double vision), loss of color vision, nystagmus

(involuntary movement of the eyes), and blurred vision.20, 21 Within the first year of disease onset, 55% of NARCOMS participants reported some degree of a visual disability.14

Pain is another frequent symptom of MS. Among the NARCOMS participants,

59% reported some degree of pain immediately after MS onset.14 Specifically, central neuropathic pain and trigeminal neuralgia are among the most common neuropathic pain syndromes experienced by MS patients.22 In addition, pain often accompanies sensory,

19 motor, and visual symptoms. For example, many MS patients suffer from painful sensory symptoms such as burning, aching, pricking, and electric-shock sensations.15 Also, MS patients often describe spasms as feeling similar to a cramping and pulling pain.22 Those

MS patients who are wheelchair users may suffer from pain due to improper posture, incorrect use of their mobility aid, and persistent wheelchair use.22 Furthermore, MS patients suffering from optic neuritis often experience eye pain, especially when they move their eyes.19

Other symptoms that MS patients may experience include loss of cognition such as memory loss and dementia,13, 23 speech and swallowing difficulties,24 bladder complications including urgency and urinary infections,25 and sexual dysfunction such as loss of interest in sex and failure of arousal.26 Overall, many patients feel extremely anxious and uncertain about their symptoms, especially once diagnosed with MS.27

2.1.2 Diagnosis

Correctly diagnosing MS is particularly challenging. For example, some symptoms of MS, such as fatigue and sensory disturbances, are non-specific to MS, and therefore may be associated with other chronic diseases.28 In addition, there is no single clinical feature or laboratory test that is entirely sufficient to make an MS diagnosis.29

However, various sets of diagnostic criteria have been developed throughout the years to support a clinical diagnosis of MS, including the Schumacher criteria,30 the Poser criteria,31 and the McDonald criteria.32 In general, these criteria are based on the following principles: (1) dissemination in space of CNS lesions, (2) dissemination in time

20 of CNS lesions, and (3) exclusion of other similar diseases that may have caused the symptoms.29

These principles were first introduced in the Schumacher criteria,30 the first official standardized set of criteria developed in 1965 for diagnosing definite MS in a purely clinical manner. In general, a clinical diagnosis of definite MS can be made using the Schumacher criteria if two separate attacks (dissemination in time) are located in two different areas of the CNS (dissemination in space), and the signs and symptoms cannot be better explained by a different disease.

The Schumacher criteria were modified and a more advanced set of diagnostic criteria, known as the Poser criteria,31 were published in 1983. These criteria incorporate laboratory tests (evoked potentials and cerebrospinal fluid analysis) to locate asymptomatic CNS lesions. Furthermore, the Poser criteria allow for a diagnosis of probable MS in addition to a diagnosis of CDMS that was already included in the

Schumacher criteria. Overall, these advancements aid in further confirming the diagnosis of MS.

The most recent set of MS diagnostic criteria, the McDonald criteria,32 were published in 2001. While the McDonald criteria still focus on demonstrating the MS diagnostic principles, they rely on magnetic resonance imaging in addition to evoked potentials and cerebrospinal fluid analysis to facilitate an earlier MS diagnosis. This advancement also assists in diagnosing patients that exhibit various MS subtypes

(described in the section that follows). Furthermore, the terms CDMS and probable MS were replaced by the terms MS and possible MS. In an effort to improve accuracy and efficiency in the diagnosis of MS, the McDonald criteria were revised in 200533 and again

21 in 2010.34 According to a global survey conducted by the World Health Organization and the MS International Federation, the Schumacher, Poser, and McDonald criteria have all been used in recent years to diagnose MS.35

2.1.3 Types

Prior to being diagnosed with MS, approximately 85% of MS patients would have suffered from CIS, which is the first clinical episode of neurological symptoms due to inflammatory demyelination of the CNS.36 While CIS is a disease course suggestive of

MS, the syndrome may or may not develop into MS. Specifically, the risk of developing

MS is high if brain lesions are found on a magnetic resonance imaging scan of a CIS patient.36

Besides CIS, the International Advisory Committee on Clinical Trials of MS outlined other MS phenotypes, including relapsing-remitting multiple sclerosis (RRMS), secondary-progressive multiple sclerosis (SPMS), and primary-progressive multiple sclerosis (PPMS).37

Most patients present with RRMS, which represents approximately 85% of MS cases.1 RRMS is characterized by episodes of symptom exacerbation and recovery.38

These periods of relapse and remission can continue for many months or years without disease progression.39 Eventually, about 80% of RRMS patients will develop SPMS.1

Specifically, following the RRMS course, a patient may experience a gradual increase in disability either with or without exacerbations.37 Lastly, a less common type of MS is

PPMS, representing 10-15% of all MS cases.1, 37, 38 From the onset, PPMS patients will experience a gradual increase in disability without prior exacerbations.37

2.1.4 Impact

Once diagnosed with a specific MS type, living with the symptoms of MS can have a negative impact on patients and their families, as well as on society as a whole.

Specifically, the symptoms can adversely affect the quality of life of patients and family members, including their physical, psychological, social, and economic well-being.40

Many MS patients who suffer from mobility problems, spasticity, instability, or fatigue experience a pronounced decline in daily physical functioning.41 In addition, pain and sensory disturbances have been classified as physically disabling symptoms given that such symptoms can impair one’s ability to undertake daily activities at work and at home.42 Furthermore, difficulty in participating in social activities and maintaining relationships is common among MS patients, especially if suffering from severe pain, limited mobility, fatigue, or a severe disability.43-45 MS patients may also choose to isolate themselves from social activities if they suffer from bladder control problems to avoid embarrassment.44 Not only do the symptoms of MS place severe limitations on social participation, they can also adversely impact one’s sexual relationship with a spouse. For instance, MS patients may demonstrate decreased affection and intimacy towards their spouse, as well as dissatisfaction with sexual activity.46

MS patients may also find it difficult to maintain employment due to cognitive impairment, mobility problems, fatigue, loss of vision, or accumulating disability.44, 47, 48 It has been reported that approximately 80% of MS patients face unemployment within 10 years of being diagnosed.35 Finally, the presence of depression and anxiety among MS patients is commonly reported in the literature, especially among those patients who suffer from a severe disability or have functional limitations.49-51

The various issues that accompany an MS diagnosis can impact the health and well-being of family members charged with caring for an MS patient. One study showed that a majority of caregivers experience back problems, fatigue, and general worry about their own health.43 In addition, a large proportion of the family members in the study reported that the patients’ illnesses influenced their careers. These influences included declining a job promotion, opting for part-time employment or a lower-grade job, or quitting their job in order to find more time to assist their relative with MS. Furthermore, family caregivers who experience disruptions in their daily activities, financial problems, or long-hours of caregiving can be at risk of developing depression and anxiety.52 Several studies have shown that the greater the disability of MS patients, the greater the caregiver burden.43, 52, 53

The economic burden of MS on patients, families, the health care system, and

Canadian society in general is extremely high. Based on data collected by the multinational Treatment Experience, Burden, and Unmet Needs study, the mean annual cost for each MS patient in Canada was $37,672 in 2009 dollars.54 This amount was attributable to direct healthcare costs (e.g., professional care and consultations), therapies, and indirect costs of productivity loss, including a patient’s sick leave and retirement due to MS. Greater disease severity led to higher costs. Another study estimated the lifetime cost of MS for one patient in Canada to be approximately $1.6 million in 1995 dollars.55

According to the Public Health Agency of Canada, the estimated total cost of MS was approximately $1 billion in 2000-2001, which included approximately $139 million in direct costs (e.g., hospital care, physician care, and drug costs) and $811 million in indirect costs (e.g., mortality and morbidity costs).56

2.1.5 Etiology

Despite the various symptoms, disease courses, and adverse effects that are associated with the diagnosis of MS, the exact cause of MS has yet to be discovered.35

Over the years, researchers have identified a number of different risk factors that could be potential causes of MS. These include genetics, viruses, infections, immune factors, vascular disturbances, psychological aspects, environmental triggers, and lifestyle behaviours.39 Currently, factors such as the Human Leukocyte Antigen (HLA) genetic complex,57, 58 the Epstein-Barr virus (EBV),59, 60 Vitamin D deficiency,61, 62 Chronic

Cerebrospinal Venous Insufficiency (CCSVI),63, 64 and smoking65, 66 are attracting considerable attention from researchers and scientists. Other risk factors that have been investigated include psychological stress,67, 68 organic solvents,69, 70 dietary fat intake,11, 71 the Hepatitis B vaccination,72, 73 and dog exposure.74, 75

MS is likely a multi-factorial disease process involving an interaction between a genetic predisposition and multiple environmental factors, which can be illustrated by

Rothman’s Model of Necessary and Sufficient Causes previously presented in Figure 1-1

(Chapter 1). This interaction may be responsible for triggering an autoimmune process that eventually leads to MS.76 As previously stated, MS usually develops in adulthood, which may be an indication that an environmental component is required to trigger disease onset later in life.39 Numerous studies provide evidence that the interaction of risk factors could increase the risk of MS, such as when HLA genes interact with EBV antibodies77 and cigarette smoke.78

2.1.5.1 Smoking

In 2014, approximately 5.8 trillion cigarettes were smoked globally.79 Nationally representative survey data from 187 countries showed that the prevalence of daily smoking is approximately five times higher in men (31.1%) compared to women (6.2%), with the highest prevalence observed at 45-49 years of age among men and 50-54 years of age among women.80 According to the World Lung Foundation and the American

Cancer Society, the prevalence of cigarette consumption is the highest in the European

Region (e.g., Russia, Germany, Turkey), followed by the Americas (e.g., United States,

Canada, Argentina), South-East Asia (e.g., Indonesia, India, Singapore), the Western

Pacific (e.g., China, Japan, Vietnam), and finally, the Eastern Mediterranean and Africa.79

In Canada, 21.4% of men and 14.8% of women reported being smokers in 2014, down from 24.2% and 17.4% reported in each respective group in 2010.81

In the 1960s, case-control studies and surveys began to investigate the relationship between smoking and MS. Results from these early studies were mixed. For instance, a nationwide case-control study in Israel showed a larger proportion of MS patients had smoked prior to MS onset compared to the control group.82 Conversely, a national survey in Britain demonstrated no relationship between smoking and MS.83 The past decade has witnessed a dramatic increase in the number of studies examining the association between smoking and MS.65, 66, 84

Numerous cohort and case-control studies demonstrated evidence that smoking is associated with MS susceptibility. For example, two cohort studies of American women showed a 60% greater MS rate among current smokers than among individuals who never smoked.85 This study also showed that the MS rate increased with cumulative

26 smoking exposure, with the greatest risk (1.7 times) among those who had smoked for 25 years or more. The large prospective design of this study confirms that smoking exposure was evaluated prior to an MS diagnosis, establishing a temporal relationship. In addition, a nested case-control study verified a higher risk of MS among current smokers and ex- smokers compared to never-smokers.84 Given that the study used a sample taken from a well-defined prospective cohort, they were also able to establish temporality, as well as minimal bias in the selection of participants. Furthermore, two Swedish case-control studies showed that an increase in smoking dose resulted in increased risk of MS; individuals classified as heavy smokers had a greater MS risk compared to light smokers.86 The study also showed that after a decade of smoking cessation, the risk of

MS subsided, regardless of timing and cumulative dose of smoking. However, these results should be taken with caution given that smoking history was collected retrospectively, making the study prone to recall errors, and the presence of a large proportion of non-respondent controls may have introduced selection bias.

There have been mixed results from studies investigating the relationship between different forms of second-hand smoke exposure and MS risk. For instance, a case-control study in France showed a relationship between exposure to household parental smoke and increased MS risk among children.87 However, a Canadian case-control study did not demonstrate a relationship between exposure to household smoke and MS among adults.88 Similarly, while a case-control study showed an effect of maternal prenatal smoke exposure on MS risk among offspring,89 a different study using two large cohorts of female nurses did not reveal a relationship.90

Studies on the relationship between smoking and MS progression show mixed

27 results. For instance, a cohort study utilizing smoking data from medical records and interviews showed an association between smoking and the conversion from CIS to

CDMS.91 On the other hand, a different cohort study that used cotinine as a measure of smoking exposure did not show a relationship.92 In addition, a study utilizing self- reported survey data showed that ever-smokers (current and past smokers in combination) had a greater likelihood of developing progressive MS compared to never- smokers.93 This association was more pronounced among smokers who started smoking at an earlier age (15 years of age or younger) than smokers who started later. A cross- sectional and longitudinal follow-up study that used a considerably larger sample concluded that current smokers had more severe MS compared to never-smokers.94 In addition, the current smokers in the study experienced a faster progression from RRMS to

SPMS compared to those who never smoked. Furthermore, a population-based cohort study in the United Kingdom demonstrated that ex-smokers had reduced disability progression compared to current smokers, regardless of whether smoking cessation occurred before or after MS onset.95 However, a different cohort study conducted in the

Netherlands revealed no effect of smoking on disability or disease progression among

MS patients, including the development of SPMS.96 Rather than being population-based, this study cohort was hospital-based, and as such, may not have been representative of the general MS population.

The relationship between smoking and MS was examined in two systematic reviews without meta-analyses.97, 98 Based on 14 articles published no later than 2012, one of the reviews showed that smoking led to an increase in MS risk, as well as a faster disease progression.97 After examining three studies identified in a search ending in June

2014, the second review did not demonstrate a consistent relationship between smoking and an MS disease course, namely RRMS and PPMS.98 However, the authors emphasized that while the research on MS risk factors is extensive, findings were rarely published according to MS disease course.

Five systematic reviews with meta-analyses have examined the association between smoking and MS.99-103 The earlier reviews all demonstrated a statistically significant relationship between smoking and MS risk based on six studies published up to and including 2005,100 14 studies published no later than May 2010,99 and 26 studies published up to and including 2014101 respectively. Based on four studies, one of the reviews also showed that the association between smoking and risk of SPMS approached statistical significance with a RR of 1.88 and a 95% confidence interval (CI) of 0.98-

3.61.99 However, more recent studies have now been published investigating the association between smoking and the progression from RRMS to SPMS that need to be incorporated into the greater assessment of the literature. In addition, studies investigating the progression from CIS to CDMS in relation to smoking have thus far not been examined in a systematic review with a meta-analysis. Arguably, such a critical stage in the disease process also warrants a further investigation.

Out of the five systematic reviews with meta-analyses, the two most recent included 26 studies (published to around mid-2015) investigating the association between ever-smoking and MS risk.102, 103 Both reviews showed a statistically significant relationship, with ORs of 1.46 (95% CI [1.33-1.59]),102 and 1.57 (95% CI [1.50-1.64]).103

In both reviews, the risk of MS was greater among current smokers than among past smokers, and one of the reviews demonstrated a dose-response relationship between

29 cigarette smoking and MS risk.102 However, some studies directly examining the relationship between smoking and MS were missing from at least one of the reviews, and indirect studies providing appropriate smoking data that are important in the relationship were also missing.66, 84, 104 Lastly, in examining the effect of passive smoke exposure on

MS risk, one review showed a statistically significant relationship based on three studies

(OR 1.24, 95% CI [1.03-1.49]),103 while the other, also based on three studies, did not

(OR 1.12, 95% CI [0.87-1.36]).102 Given that this was the first time that the relationship between second-hand smoke exposure and MS has been examined in a systematic review and meta-analysis, there is a need to summarize the literature in this area in greater detail that includes both passive and prenatal smoke exposure. Such exposures could be critical

MS risk factors that require further investigation.

Overall, while the reviews outlined above provide important data evaluating the association between smoking and MS, the use of Hill’s criteria as a framework to determine the likely causality of this relationship would further advance the understanding of this risk factor in relation to the outcome.

Several narrative reviews have also confirmed the growing evidence supporting the association between smoking and MS.105-108 However, authors of narrative reviews may selectively cite the literature to support a preconceived argument, rather than use comprehensive procedures to identify and analyze all studies relevant to a certain research topic as in a systematic review and meta-analysis.109 As such, narrative reviews may be more prone to bias and authors may draw misleading conclusions.

Overall, there is value to conducting a new systematic review and meta-analysis that includes both more recently published and previously missed studies on smoking and

MS risk, and also examines the effect of smoking on important MS progression outcomes. Relevant outcomes include expanding on the analysis of secondary progression, as well as, for the first time in a systematic review and meta-analysis, examining the conversion from CIS to CDMS. In addition, providing a summary of the literature to date on the association between different forms of second-hand smoke exposure and MS is informative as it represents an important area of inquiry. Moreover, incorporating Hill’s criteria to determine the likelihood of a causal relationship between smoking and MS would further enhance this more timely and comprehensive review.

2.2 Overview of Systematic Review and Meta-Analytic Methods

A systematic review involves utilizing a methodical procedure to find, evaluate, and synthesize data from multiple studies.110 Prior to conducting a systematic review, it is recommended to first register a detailed review protocol. An example of a database used for registration is PROSPERO,111 which compiles registered reviews in areas such as health, welfare, education, and justice. By keeping a record of pre-registered reviews,

PROSPERO strives to help reviewers avoid duplicating an existing review, as well as to reduce the risk of reporting biases by allowing important elements highlighted in a published review and those recorded in the registered protocol to be compared.

A systematic review begins with the development of eligibility criteria (e.g., study design, nature of independent and dependent variables, population and participant characteristics) in order to determine which studies are to be included in a specific review, with the intention of finding every study satisfying the select criteria.110 From there, bibliographic databases, conference abstracts, theses or dissertations, and/or other

31 sources are searched for studies applicable to the review.112 A search for additional studies identified in the reference lists of associated studies and review articles can also be conducted. Once the studies are retrieved, they are individually screened in order to determine if they meet the eligibility criteria, and therefore, are relevant to the review.113

As such, it is imperative to consult the established criteria to ensure that they are consistently applied. In order to assist researchers in reporting their reviews, the Preferred

Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement114 was established; this includes a detailed flow diagram which, by convention, is included by many authors in their publications. Specifically, the PRISMA flow diagram allows review authors to document the number of records located in their search, screened, and excluded from the review. The full-text publications of those records not excluded are then retrieved. Subsequently, review authors can use the PRISMA flow diagram to document the number of full-text studies screened, excluded, and finally, included in the review.

Data applicable to the review (e.g., study design, participant characteristics, results) are then extracted from each included study, which may involve using a data extraction form to ensure that the procedure is undertaken as consistently as possible.113 It is essential that the form first be piloted on a small group of included studies to confirm that the form is appropriately customized to the review topic, and is efficient in obtaining all necessary data.

Another important component of a systematic review involves evaluating the risk of bias and quality of the included studies, which can include assessing for such elements as selection bias (e.g., sample representativeness, loss to follow-up, volunteer and

32 response bias) and confounder control.112 Several instruments have been identified as being helpful in facilitating such an evaluation process, including the Downs and Black instrument115 and the Newcastle-Ottawa Scale.116 Reviewers can also evaluate the quality of evidence overall using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach,117 which involves evaluating factors such as study limitations, inconsistency and imprecision of results, and publication bias.

Besides reducing bias through the use of systematic methods as presented above, systematic reviews are beneficial in terms of rectifying conflicting findings, investigating clinical practice alternatives, and verifying suitable clinical practice to assist individuals

(e.g., health professionals, healthcare users) in making sensible healthcare choices in the face of an overwhelming amount of health data.112 In terms of synthesizing the results, study findings can be summarized by means of a narrative synthesis, or, if results are displayed quantitatively and studies are relatively similar, results from individual studies can be combined using a meta-analysis.113

A meta-analysis involves utilizing statistical procedures to integrate the results of separate studies; these results (e.g., individual study effect estimates and summary meta- analytic effect estimate with their corresponding CIs) are graphically displayed in forest plots accompanied by a listing of the included studies.118 Such a statistical analysis is beneficial as it can contribute to increased precision of effect estimates and higher power as a result of establishing a larger total sample size.119

A meta-analysis can be performed using either a fixed-effects or a random-effects model.110 Specifically, a fixed-effects model assumes that one true effect is present in every study, which is estimated by the summary effect.110 Therefore, the only source of

33 error in this model is random sampling variation. A fixed-effects model can be conducted if it is assumed that the included studies are the same (e.g., same researchers, one pool of participants), with the intention of estimating the effect strictly for the population under study as opposed to making broad generalizations.110 However, since the results are usually obtained from independent studies, it is likely that the studies vary in their design and other characteristics, and such differences are likely to have influenced the results.

As such, an assumption of one true effect should not be made. Instead, a random-effects model can be conducted, which assumes that a distribution of true effects is present rather than one effect, with the intention of extrapolating to different circumstances (e.g., populations).110 The summary effect estimate represents the mean of the individual effects. Under this model, the two sources of error include within-study sampling variation and between-study variation in the true effects.110

An odds ratio (OR) is a common measure of the effect of an exposure on a given disease outcome.120 Specifically, the OR signifies the odds of a given disease developing in the presence of an exposure in comparison to the odds of the disease developing without the exposure. The OR is calculated using the following formula:

Outcome

+ - OR = a/c + b/d a b Exposure - c d OR = exposed cases/non-exposed cases exposed controls/non-exposed controls

An OR > 1 signifies that an exposure increases the odds of a given disease outcome, an OR < 1 signifies that an exposure decreases the odds of the disease, and an

OR = 1 signifies that an exposure has no effect on the odds of the disease.121 In a similar context, a risk ratio (RR=(a/a+b)/(c/c+d)) signifies the risk of an outcome in an exposed group in comparison to the risk of the outcome in the non-exposed group.112

Traditionally, RRs were calculated for cohort studies and ORs for case-control studies.

However, ORs are now commonly used with various study designs involving dichotomous outcomes.

Finally, a hazard ratio (HR) is an effect measure that is commonly used to analyze time-to-event outcomes (e.g., time to symptom appearance, time to death, etc.).121

Specifically, a HR considers both the number of events and their timing, which are critical when analyzing such outcomes. On the other hand, an OR and RR only consider the number of events, and therefore, such effect measures should not be used to investigate time-to-event outcomes.

In summary, systematic reviews and meta-analyses comprise comprehensive research methods whereby all studies relevant to a particular research question can be identified, assessed, and synthesized (quantitatively and/or qualitatively), providing reliable and precise results that can be confidently used to draw important conclusions.

2.3 Previous Use of Hill’s Criteria

2.3.1 Previous Use of Hill’s Criteria in Narrative Reviews of MS Causation

Several narrative reviews have applied the Bradford Hill criteria for causation as a framework to assess the causal nature of risk factors in relation to MS, including Vitamin

D insufficiency122 and the Hepatitis B vaccine.123 After applying Hill’s criteria, the authors of these reviews reported that both relationships may be causal. Specifically, it

35 was reported that the findings available from the literature on the relationship between

Vitamin D insufficiency and MS fulfilled all of Hill’s criteria with the exception of

Experimental Evidence; experimental evidence demonstrating whether optimizing

Vitamin D levels reduced MS risk was not available.122 The authors concluded that specificity between Vitamin D status and MS existed in terms of biological mechanisms involving inflammatory pathways in general, despite Vitamin D insufficiency being linked to many other diseases. On the other hand, the authors examining findings from the literature on the Hepatitis B vaccine and MS reported that the association fulfilled all of Hill’s criteria except Biological Gradient and Specificity; the relationship did not follow a dose-response curve, and it was non-specific given that other risk factors are likely involved in the development of MS (besides Hepatitis B).123 The authors argued that experimental evidence supporting the relationship between the Hepatitis B vaccine and MS was available; however, the evidence was based on studies that investigated immune responses generated in immunized animals rather than humans.

The Bradford Hill criteria for causation have also been used to evaluate the relationship between smoking and MS in two narrative reviews.106, 108 An earlier review was based on approximately 50 studies published up to and including 2011, including case-control and cohort studies, as well as systematic review and meta-analysis data.108

Based on these studies, the author reported that the effect of smoking on MS was small, consistent across different populations and study designs, non-specific, plausible, and coherent. On the other hand, the Experimental Evidence criterion could not be applied given that randomized controlled trials of smoking and MS were unavailable (and would be unethical to conduct). The Analogy criterion was also regarded as weak. In addition,

36 methodological problems across studies made it challenging to evaluate Temporality, as well as compromised the validity of data needed to evaluate Biological Gradient—despite many studies providing evidence of a dose-response relationship. Overall, the author concluded that while smoking was associated with MS risk, as well as disease progression and disability, more research was needed to provide further support for the causal nature of the smoking-MS relationship. On the other hand, the more recent narrative review included original studies and review articles published up to September

2014 that investigated many MS risk factors, including approximately 10 studies related to smoking.106 From these studies, the authors reported that the smoking-MS relationship was weak, consistent, plausible, and coherent, and exhibited a temporal and dose- response association. Overall, the authors concluded that smoking was associated with increased MS risk, while its impact on disease activity was more uncertain. The authors emphasized the need for more evidence to support causation.

Unfortunately, a major issue with narrative reviews is the ability of authors to be selective in reporting studies that support their preconceived views surrounding a relationship,109 therefore allowing the possibility that authors would draw misleading conclusions. This limitation is addressed with a systematic review and meta-analysis—an approach utilizing methodical procedures for reviewing all known studies on smoking and MS in the literature in order to obtain the most reliable findings. Subsequently applying Hill’s criteria to these systematically obtained findings to comprehensively assess the causal nature of the smoking-MS relationship thus further enhances the understanding generated by the review.

2.3.2 Previous Use of Hill’s Criteria in Systematic Reviews and Meta-Analyses

The Bradford Hill criteria for causation have been used in systematic reviews and meta-analyses to evaluate whether an association is causal, including the relationship between dietary factors and coronary heart disease,124 and the association between knee loading and knee osteoarthritis,125 but not the relationship between smoking and MS.

Whether or not each criterion was satisfied was based on pre-established parameters specific to the reviews. To determine the Strength of the associations, both reviews examined the magnitude, direction, and statistical significance of the pooled effect estimates (OR or RR). In addition, replication of strong or moderate associations across studies provided the authors with evidence of Consistency. Temporality was also established given that prospective cohort studies were included in the analyses. Evidence of a Biological Gradient was determined by examination of tests for a trend in the relationship between the exposures and outcomes in question. To establish Coherence, the authors of the diet/coronary heart disease review stated that there needed to be evidence supporting an association of dietary exposures with surrogate risk factors for cardiovascular disease outcomes, as well as with indicators of such outcomes.124 With regards to Experimental Evidence, randomized controlled trials were analyzed for evidence of statistically significant associations between dietary exposures and coronary outcomes.

While the Specificity, Plausibility, and Analogy criteria were not evaluated in these two reviews, the authors of the diet/coronary heart disease review discussed how the dietary exposures would be associated with many cardiovascular disease outcomes given the interconnectedness of these outcomes, making it a non-specific relationship.124

In terms of Plausibility, it was stated from the onset that the relationship met the criterion given that credible mechanisms existed that explained the relationship. Furthermore, the authors of the knee loading/osteoarthritis review noted the possibility of associations existing that were analogous to the relationship between knee loading and knee osteoarthritis, such as hip or hand osteoarthritis.125 Overall, Hill’s criteria were used by these reviews to develop a causation score for the association in question, illustrating whether there was strong, moderate, or weak evidence supporting a cause-and-effect relationship.124, 125 Employing a similar methodology to evaluate smoking as an MS risk factor would greatly assist in determining if smoking is playing a causal role in this chronic disease.

2.4 Summary

MS is a neurological condition with a high burden of disease in Canada, and is an important issue on a global scale. Especially concerning is the unpredictable nature of the disease, with its variable symptoms and disease courses, as well as challenging diagnosis, and substantial impact on patients, family members, the health care system, and society in general. While the cause of MS is unknown, numerous observational studies and several systematic reviews and meta-analyses have established that smoking is a risk factor for the disease. However, recent systematic reviews and meta-analyses failed to include some studies relevant to the association between smoking and MS risk. Systematic review and meta-analysis research examining important MS progression outcomes in relation to smoking is also lacking, including only one dated review investigating the relationship between smoking and the conversion from RRMS to SPMS and none

39 examining the association between smoking and the conversion from CIS to CDMS. In addition, systematic reviews with meta-analyses have only recently begun to examine the relationship between passive smoke exposure and MS risk. A new, more inclusive review is warranted, specifically examining different forms of second-hand smoke exposure that could represent important risk factors in the development of MS. Finally, the Bradford

Hill criteria for causation, an important tool that can be used to decipher whether smoking is likely playing a causal role in MS, have not been optimally used in previous systematic reviews and meta-analyses. The research presented in this thesis aims to examine the association between smoking and MS through a systematic review and meta- analysis incorporating the Bradford Hill criteria for causation, with a particular focus on

MS risk including additional relevant studies not included in previously published reviews, as well as a more comprehensive focus on MS progression outcomes (i.e., CIS to CDMS, and RRMS to SPMS). The research also aims to summarize the literature on the relationship between various forms of second-hand smoke exposure (i.e., passive, prenatal) and MS within a systematic review.

2.5 References

1. Multiple Sclerosis International Federation. Atlas of MS 2013. London, UK: Multiple

Sclerosis International Federation, 2013.

2. Kurland LT, Reed D. Geographic and climatic aspects of multiple sclerosis: A review of current hypotheses. Am J Public Health Nations Health 1964;54:588-97.

3. Poppe AY, Wolfson C, Zhu B. Prevalence of multiple sclerosis in Canada: A systematic review. Can J Neurol Sci 2008;35(5):593-601.

4. Wade BJ. Spatial analysis of global prevalence of multiple sclerosis suggests need for an updated prevalence scale. Mult Scler Int 2014;2014:124578.

5. Warren SA, Svenson LW, Warren KG. Contribution of incidence to increasing prevalence of multiple sclerosis in Alberta, Canada. Mult Scler 2008;14(7):872-9.

6. Hader WJ, Yee IM. Incidence and prevalence of multiple sclerosis in Saskatoon,

Saskatchewan. Neurology 2007;69(12):1224-9.

7. Marrie RA, Yu N, Blanchard J, Leung S, Elliott L. The rising prevalence and changing age distribution of multiple sclerosis in Manitoba. Neurology 2010;74(6):465-71.

8. Marrie RA, Fisk JD, Stadnyk KJ, et al. The incidence and prevalence of multiple sclerosis in Nova Scotia, Canada. Can J Neurol Sci 2013;40(6):824-31.

9. Tullman MJ. Overview of the epidemiology, diagnosis, and disease progression associated with multiple sclerosis. Am J Manag Care 2013;19(2 Suppl):S15-20.

10. Milo R, Kahana E. Multiple sclerosis: Geoepidemiology, genetics and the environment. Autoimmun Rev 2010;9(5):A387-94.

11. Esparza ML, Sasaki S, Kesteloot H. Nutrition, latitude, and multiple sclerosis mortality: An ecologic study. Am J Epidemiol 1995;142(7):733-7.

12. Embrey N. Multiple sclerosis: Managing a complex neurological disease. Nurs Stand

2014;29(11):49-58.

13. Cosh A, Carslaw H. Multiple sclerosis: Symptoms and diagnosis. InnovAiT

2014;7(11):651-7.

14. Kister I, Bacon TE, Chamot E, et al. Natural history of multiple sclerosis symptoms.

Int J MS Care 2013;15(3):146-58.

15. Osterberg A, Boivie J. Central pain in multiple sclerosis - sensory abnormalities. Eur

J Pain 2010;14(1):104-10.

16. Alusi SH, Worthington J, Glickman S, Bain PG. A study of tremor in multiple sclerosis. Brain 2001;124(Pt 4):720-30.

17. Barnes MP, Kent RM, Semlyen JK, McMullen KM. Spasticity in multiple sclerosis.

Neurorehabil Neural Repair 2003;17(1):66-70.

18. Kasser SL, Jacobs JV, Foley JT, Cardinal BJ, Maddalozzo GF. A prospective evaluation of balance, gait, and strength to predict falling in women with multiple sclerosis. Arch Phys Med Rehabil 2011;92(11):1840-6.

19. Chan JW. Optic neuritis in multiple sclerosis. Ocul Immunol Inflamm

2002;10(3):161-86.

20. Graves J, Balcer LJ. Eye disorders in patients with multiple sclerosis: Natural history and management. Clin Ophthalmol 2010;4:1409-22.

21. Hurwitz BJ. The diagnosis of multiple sclerosis and the clinical subtypes. Ann Indian

Acad Neurol 2009;12(4):226-30.

22. Solaro C, Trabucco E, Messmer Uccelli M. Pain and multiple sclerosis:

Pathophysiology and treatment. Curr Neurol Neurosci Rep 2013;13(1):320.

23. Newland PK, Thomas FP, Riley M, Flick LH, Fearing A. The use of focus groups to characterize symptoms in persons with multiple sclerosis. J Neurosci Nurs

2012;44(6):351-7.

24. Hartelius L, Svensson P. Speech and swallowing symptoms associated with

Parkinson's disease and multiple sclerosis: A survey. Folia Phoniatr Logop 1994;46(1):9-

17.

25. Nakipoglu GF, Kaya AZ, Orhan G, et al. Urinary dysfunction in multiple sclerosis.

Journal of Clinical Neuroscience 2008;16(10):1321-4.

26. Khan F, Pallant JF, Ng L, Whishaw M. Sexual dysfunction in multiple sclerosis. Sex

Disabil 2011;29(2):101-11.

27. Isaksson AK, Ahlstrom G. From symptom to diagnosis: Illness experiences of multiple sclerosis patients. J Neurosci Nurs 2006;38(4):229-37.

28. Palace J. Making the diagnosis of multiple sclerosis. J Neurol Neurosurg Psychiatry

2001;71 Suppl 2:ii3-8.

29. Milo R, Miller A. Revised diagnostic criteria of multiple sclerosis. Autoimmun Rev

2014;13(4-5):518-24.

30. Schumacher GA, Beebe G, Kibler RF, et al. Problems of experimental trials of therapy in multiple sclerosis: Report by the Panel on the Evaluation of Experimental

Trials of Therapy in Multiple Sclerosis. Ann N Y Acad Sci 1965;122:552-68.

31. Poser CM, Paty DW, Scheinberg L, et al. New diagnostic criteria for multiple sclerosis: Guidelines for research protocols. Ann Neurol 1983;13(3):227-31.

32. McDonald WI, Compston A, Edan G, et al. Recommended diagnostic criteria for multiple sclerosis: Guidelines from the International Panel on the Diagnosis of Multiple

Sclerosis. Ann Neurol 2001;50(1):121-7.

33. Polman CH, Reingold SC, Edan G, et al. Diagnostic criteria for multiple sclerosis:

2005 revisions to the "McDonald Criteria". Ann Neurol 2005;58(6):840-6.

34. Polman CH, Reingold SC, Banwell B, et al. Diagnostic criteria for multiple sclerosis:

2010 revisions to the McDonald Criteria. Ann Neurol 2010;69:292-302.

35. World Health Organization. Atlas: Multiple sclerosis resources in the world. Geneva,

CH: WHO Press, 2008.

36. Miller D, Barkhof F, Montalban X, Thompson A, Filippi M. Clinically isolated syndromes suggestive of multiple sclerosis, part I: Natural history, pathogenesis, diagnosis, and prognosis. Lancet Neurol 2005;4(5):281-8.

37. Lublin FD, Reingold SC, Cohen JA, et al. Defining the clinical course of multiple sclerosis: The 2013 revisions. Neurology 2014;83(3):278-86.

38. Murray TJ. Multiple sclerosis: The history of a disease. New York, NY: Demos

Medical Publishing, 2005.

39. Talley CL. A history of multiple sclerosis. Westport, CT: Praeger, 2008.

40. Jones CA, Pohar SL, Warren S, Turpin KV, Warren KG. The burden of multiple sclerosis: A community health survey. Health Qual Life Outcomes 2008;6:1.

41. Isaksson AK, Ahlstrom G, Gunnarsson LG. Quality of life and impairment in patients with multiple sclerosis. J Neurol Neurosurg Psychiatry 2005;76(1):64-9.

42. Beiske AG, Pedersen ED, Czujko B, Myhr KM. Pain and sensory complaints in multiple sclerosis. Eur J Neurol 2004;11(7):479-82.

43. Hakim EA, Bakheit AM, Bryant TN, et al. The social impact of multiple sclerosis--a study of 305 patients and their relatives. Disabil Rehabil 2000;22(6):288-93.

44. Halper J. The psychosocial effect of multiple sclerosis: The impact of relapses. J

Neurol Sci 2007;256 Suppl 1:S34-8.

45. Kalia LV, O'Connor PW. Severity of chronic pain and its relationship to quality of life in multiple sclerosis. Mult Scler 2005;11(3):322-7.

46. Esmail S, Munro B, Gibson N. Couple's experience with multiple sclerosis in the context of their sexual relationship. Sex Disabil 2007;25(4):163-77.

47. Benedict RH, Wahlig E, Bakshi R, et al. Predicting quality of life in multiple sclerosis: Accounting for physical disability, fatigue, cognition, mood disorder, personality, and behavior change. J Neurol Sci 2005;231(1-2):29-34.

48. Julian LJ, Vella L, Vollmer T, Hadjimichael O, Mohr DC. Employment in multiple sclerosis. Exiting and re-entering the work force. J Neurol 2008;255(9):1354-60.

49. Janssens AC, van Doorn PA, de Boer JB, van der Meche FG, Passchier J, Hintzen

RQ. Impact of recently diagnosed multiple sclerosis on quality of life, anxiety, depression and distress of patients and partners. Acta Neurol Scand 2003;108(6):389-95.

50. Lester K, Stepleman L, Hughes M. The association of illness severity, self-reported cognitive impairment, and perceived illness management with depression and anxiety in a multiple sclerosis clinic population. J Behav Med 2007;30(2):177-86.

51. Tsivgoulis G, Triantafyllou N, Papageorgiou C, et al. Associations of the Expanded

Disability Status Scale with anxiety and depression in multiple sclerosis outpatients. Acta

Neurol Scand 2006;115:67-72.

52. Buhse M. Assessment of caregiver burden in families of persons with multiple sclerosis. J Neurosci Nurs 2008;40(1):25-31.

53. Buchanan RJ, Radin D, Huang C. Caregiver burden among informal caregivers assisting people with multiple sclerosis. Int J MS Care 2011;13(2):76-83.

54. Karampampa K, Gustavsson A, Miltenburger C, Kindundu CM, Selchen DH.

Treatment experience, burden, and unmet needs (TRIBUNE) in multiple sclerosis: The costs and utilities of MS patients in Canada. J Popul Ther Clin Pharmacol

2012;19(1):e11-25.

55. Canadian Burden of Illness Study Group. Burden of illness of multiple sclerosis: Part

I: Cost of illness. Can J Neurol Sci 1998;25(1):23-30.

56. Canadian Institute for Health Information. The burden of neurological diseases, disorders, and injuries in Canada. Ottawa, ON: CIHI, 2007.

57. Abolfazli R, Samadzadeh S, Sabokbar T, et al. Relationship between HLA-DRB1*

11/15 genotype and susceptibility to multiple sclerosis in Iran. J Neurol Sci 2014;345(1-

2):92-6.

58. Sawcer S, Franklin RJ, Ban M. Multiple sclerosis genetics. Lancet Neurol

2014;13(7):700-9.

59. Almohmeed YH, Avenell A, Aucott L, Vickers MA. Systematic review and meta- analysis of the sero-epidemiological association between Epstein Barr virus and multiple sclerosis. PLoS One 2013;8(4):e61110.

60. Pakpoor J, Giovannoni G, Ramagopalan SV. Epstein-Barr virus and multiple sclerosis: Association or causation? Expert Rev Neurother 2013;13(3):287-97.

61. Harandi AA, Pakdaman H, Sahraian MA. Vitamin D and multiple sclerosis. Iran J

Neurol 2014;13(1):1-6.

62. Pandit L, Ramagopalan SV, Malli C, D'Cunha A, Kunder R, Shetty R. Association of vitamin D and multiple sclerosis in India. Mult Scler 2013;19(12):1592-6.

63. Traboulsee AL, Knox KB, Machan L, et al. Prevalence of extracranial venous narrowing on catheter venography in people with multiple sclerosis, their siblings, and unrelated healthy controls: A blinded, case-control study. Lancet 2014;383(9912):138-45.

64. Zwischenberger BA, Beasley MM, Davenport DL, Xenos ES. Meta-analysis of the correlation between chronic cerebrospinal venous insufficiency and multiple sclerosis.

Vasc Endovascular Surg 2013;47(8):620-4.

65. O'Gorman C, Bukhari W, Todd A, Freeman S, Broadley SA. Smoking increases the risk of multiple sclerosis in Queensland, Australia. J Clin Neurosci 2014;21(10):1730-3.

66. Salzer J, Hallmans G, Nystrom M, Stenlund H, Wadell G, Sundstrom P. Smoking as a risk factor for multiple sclerosis. Mult Scler 2013;19(8):1022-7.

67. Riise T, Mohr DC, Munger KL, Rich-Edwards JW, Kawachi I, Ascherio A. Stress and the risk of multiple sclerosis. Neurology 2011;76(22):1866-71.

68. Warren SA, Warren KG, Greenhill S, Paterson M. How multiple sclerosis is related to animal illness, stress and diabetes. Can Med Assoc J 1982;126(4):377-82, 385.

69. Landtblom AM, Flodin U, Soderfeldt B, Wolfson C, Axelson O. Organic solvents and multiple sclerosis: A synthesis of the current evidence. Epidemiology 1996;7(4):429-

33.

70. Mortensen JT, Bronnum-Hansen H, Rasmussen K. Multiple sclerosis and organic solvents. Epidemiology 1998;9(2):168-171.

71. Zhang SM, Willett WC, Hernan MA, Olek MJ, Ascherio A. Dietary fat in relation to risk of multiple sclerosis among two large cohorts of women. Am J Epidemiol

2000;152(11):1056-64.

72. Ascherio A, Zhang SM, Hernan MA, et al. Hepatitis B vaccination and the risk of multiple sclerosis. N Engl J Med 2001;344(5):327-32.

73. DeStefano F, Verstraeten T, Chen RT. Hepatitis B vaccine and risk of multiple sclerosis. Expert Rev Vaccines 2002;1(4):461-6.

74. Cook SD, Natelson BH, Levin BE, Chavis PS, Dowling PC. Further evidence of a possible association between house dogs and multiple sclerosis. Ann Neurol

1978;3(2):141-3.

75. Kurtzke JF, Priester WA. Dogs, distemper, and multiple sclerosis in the United

States. Acta Neurol Scand 1979;60(5):312-9.

76. Jankosky C, Deussing E, Gibson RL, Haverkos HW. Viruses and vitamin D in the etiology of type 1 diabetes mellitus and multiple sclerosis. Virus Res 2012;163(2):424-30.

77. Sundqvist E, Sundstrom P, Linden M, et al. Epstein-Barr virus and multiple sclerosis:

Interaction with HLA. Genes Immun 2012;13(1):14-20.

78. Hedstrom AK, Bomfim IL, Barcellos LF, et al. Interaction between passive smoking and two HLA genes with regard to multiple sclerosis risk. Int J Epidemiol

2014;43(6):1791-8.

79. Eriksen M, Mackay J, Schluger N, Gomeshtapeh FI, Drope J. The tobacco atlas.

Atlanta, GA: American Cancer Society, 2015.

80. Ng M, Freeman MK, Fleming TD, et al. Smoking prevalence and cigarette consumption in 187 countries, 1980-2012. JAMA 2014;311(2):183-92.

81. Statistics Canada. Smokers, by Sex, Provinces and Territories (Percent). CANSIM.

Catalogue no. 82-221-X. Ottawa, ON: Statistics Canada, 2016.

82. Antonovsky A, Leibowitz U, Smith HA, et al. Epidemiologic study of multiple sclerosis in Israel. I. An overall review of methods and findings. Arch Neurol

1965;13:183-93.

83. Simpson CA, Newell DJ, Schapira K. Smoking and multiple sclerosis. Neurology

1966;16(10):1041-3 passim.

84. Hernan MA, Jick SS, Logroscino G, Olek MJ, Ascherio A, Jick H. Cigarette smoking and the progression of multiple sclerosis. Brain 2005;128(Pt 6):1461-5.

85. Hernan MA, Olek MJ, Ascherio A. Cigarette smoking and incidence of multiple sclerosis. Am J Epidemiol 2001;154(1):69-74.

86. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Smoking and multiple sclerosis susceptibility. Eur J Epidemiol 2013;28(11):867-74.

87. Mikaeloff Y, Caridade G, Tardieu M, Suissa S. Parental smoking at home and the risk of childhood-onset multiple sclerosis in children. Brain 2007;130(Pt 10):2589-95.

88. Ramagopalan SV, Lee JD, Yee IM, et al. Association of smoking with risk of multiple sclerosis: A population-based study. J Neurol 2013;260(7):1778-81.

89. Mueller BA, Nelson JL, Newcomb PA. Intrauterine environment and multiple sclerosis: A population- based case-control study. Mult Scler 2013;19(1):106-11.

90. Gardener H, Munger KL, Chitnis T, Michels KB, Spiegelman D, Ascherio A.

Prenatal and perinatal factors and risk of multiple sclerosis. Epidemiology

2009;20(4):611-8.

91. Di Pauli F, Reindl M, Ehling R, et al. Smoking is a risk factor for early conversion to clinically definite multiple sclerosis. Mult Scler 2008;14(8):1026-30.

92. Horakova D, Zivadinov R, Weinstock-Guttman B, et al. Environmental factors associated with disease progression after the first demyelinating event: Results from the multi-center SET study. PLoS One 2013;8(1):e53996.

93. Sundstrom P, Nystrom L. Smoking worsens the prognosis in multiple sclerosis. Mult

Scler 2008;14(8):1031-5.

94. Healy BC, Ali EN, Guttmann CR, et al. Smoking and disease progression in multiple sclerosis. Arch Neurol 2009;66(7):858-64.

95. Manouchehrinia A, Tench CR, Maxted J, Bibani RH, Britton J, Constantinescu CS.

Tobacco smoking and disability progression in multiple sclerosis: United Kingdom cohort study. Brain 2013;136(Pt 7):2298-304.

96. Koch M, van Harten A, Uyttenboogaart M, De Keyser J. Cigarette smoking and progression in multiple sclerosis. Neurology 2007;69(15):1515-20.

97. McKay KA, Jahanfar S, Duggan T, Tkachuk S, Tremlett H. Factors associated with onset, relapses or progression in multiple sclerosis: A systematic review. Neurotoxicology

2016 (in press).

98. McKay KA, Kwan V, Duggan T, Tremlett H. Risk factors associated with the onset of relapsing-remitting and primary progressive multiple sclerosis: A systematic review.

Biomed Res Int 2015;2015:817238.

99. Handel AE, Williamson AJ, Disanto G, Dobson R, Giovannoni G, Ramagopalan SV.

Smoking and multiple sclerosis: An updated meta-analysis. PLoS One 2011;6(1):e16149.

100. Hawkes CH. Smoking is a risk factor for multiple sclerosis: A metanalysis. Mult

Scler 2007;13(5):610-5.

101. O'Gorman C, Broadley SA. Smoking and multiple sclerosis: Evidence for latitudinal and temporal variation. J Neurol 2014;261(9):1677-83.

102. Poorolajal J, Bahrami M, Karami M, Hooshmand E. Effect of smoking on multiple sclerosis: A meta-analysis. J Public Health (Oxf) 2016. doi:10.1093/pubmed/fdw030.

103. Zhang P, Wang R, Li Z, et al. The risk of smoking on multiple sclerosis: A meta- analysis based on 20,626 cases from case-control and cohort studies. PeerJ

2016;4:e1797.

104. Moccia M, Lanzillo R, Palladino R, et al. The Framingham cardiovascular risk score in multiple sclerosis. Eur J Neurol 2015;22(8):1176-83.

105. Jafari N, Hintzen RQ. The association between cigarette smoking and multiple sclerosis. J Neurol Sci 2011;311(1-2):78-85.

106. Loken-Amsrud KI, Lossius A, Torkildsen O, Holmoy T. Impact of the environment on multiple sclerosis. Tidsskr Nor Laegeforen 2015;135(9):856-60.

107. Shirani A, Tremlett H. The effect of smoking on the symptoms and progression of multiple sclerosis: A review. J Inflamm Res 2010;3:115-26.

108. Wingerchuk DM. Smoking: Effects on multiple sclerosis susceptibility and disease progression. Ther Adv Neurol Disord 2012;5(1):13-22.

109. Garg AX, Hackam D, Tonelli M. Systematic review and meta-analysis: When one study is just not enough. Clin J Am Soc Nephrol 2008;3(1):253-60.

110. Borenstein M, Hedges LV, Higgins JP, Rothstein HR. Introduction to meta- analysis. Chichester, UK: Wiley, 2009.

111. National Institute for Health Research. PROSPERO: International prospective register of systematic reviews. n.d. Available at: http://www.crd.york.ac.uk/PROSPERO/about.php?about=about.

112. Higgins JPT, Green S (editors). Cochrane handbook for systematic reviews of interventions (version 5.1.0) [updated March 2011]. Cochrane Collaboration, 2011.

Available at: www.handbook.cochrane.org.

113. Denison HJ, Dodds RM, Ntani G, et al. How to get started with a systematic review in epidemiology: An introductory guide for early career researchers. Arch Public Health

2013;71(1):21.

114. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. PLoS Med

2009;6(7):e1000097.

115. Downs SH, Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions. J Epidemiol Community Health 1998;52(6):377-84.

116. Wells GA, Shea B, O'Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses. 2014. Available at: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp.

117. Guyatt GH, Oxman AD, Kunz R, Vist GE, Falck-Ytter Y, Schunemann HJ. What is

"quality of evidence" and why is it important to clinicians? BMJ 2008;336(7651):995-8.

118. Cooper H, Hedges LV, Valentine JC. The handbook of research synthesis and meta- analysis. New York, NY: Russell Sage Foundation, 2009.

119. Akobeng AK. Understanding systematic reviews and meta-analysis. Arch Dis Child

2005;90(8):845-8.

120. Szumilas M. Explaining odds ratios. J Can Acad Child Adolesc Psychiatry

2010;19(3):227-9.

121. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 2007;8:16.

122. Hanwell HE, Banwell B. Assessment of evidence for a protective role of vitamin D in multiple sclerosis. Biochim Biophys Acta 2011;1812(2):202-12.

123. Le Houezec D. Evolution of multiple sclerosis in France since the beginning of hepatitis B vaccination. Immunol Res 2014;60(2-3):219-25.

124. Mente A, de Koning L, Shannon HS, Anand SS. A systematic review of the evidence supporting a causal link between dietary factors and coronary heart disease.

Arch Intern Med 2009;169(7):659-69.

125. Henriksen M, Creaby MW, Lund H, Juhl C, Christensen R. Is there a causal link between knee loading and knee osteoarthritis progression? A systematic review and meta-analysis of cohort studies and randomised trials. BMJ Open 2014;4(7):e005368.

CHAPTER 3: Manuscript

Smoking and Multiple Sclerosis: A Systematic Review and Meta-Analysis Using the

Bradford Hill Criteria for Causation

Abstract

Smoking had a statistically significant association with both MS risk (conservative:

OR/RR 1.54, 95% CI [1.46-1.63]) and SPMS risk (HR 1.80, 95% CI [1.04-3.10]), but the association with progression from CIS to CDMS was non-significant (HR 1.13, 95% CI

[0.73-1.76]). Based on Hill’s criteria, there was strong evidence of a causal role of

55 smoking in MS risk, but only moderate evidence of a causal association between smoking and MS progression. Heterogeneity in study designs and target populations, inconsistent results, and an overall scarcity of studies point to the need for more research in the area of second-hand smoke exposure in relation to MS prior to conducting a detailed meta- analysis. Conclusion. This first review to supplement systematic review and meta- analytic methods with the application of Hill’s criteria to analyze the smoking-MS association provided evidence supporting the causal involvement of smoking in the development and progression of MS. As such, smoking cessation programs and policies should highlight MS as an additional health risk to deter the public from smoking.

Keywords: Multiple sclerosis; cigarette smoking; clinically isolated syndrome; Bradford

Hill

56 1. Introduction

Multiple sclerosis (MS) is among the most common neurological diseases in the world.1 The disease affected approximately 2.3 million people worldwide in 2013, with the highest prevalence reported in North America and Europe.1 MS is characterized by damage to the myelin surrounding the nerves of the central nervous system (CNS), as well as damage to the nerve axons, contributing to a disruption in impulse conduction which ultimately leads to MS symptoms.2 The cause of MS remains unknown.3 However,

MS likely involves an interaction between genetic and environmental factors.4 Currently, the Human Leukocyte Antigen (HLA),5, 6 Epstein-Barr virus (EBV),7, 8 Vitamin D deficiency,9, 10 and cigarette smoking11, 12 are thought to be involved.

The past decade witnessed an increase in the number of studies examining the association between smoking and MS. Numerous studies have shown that smoking is associated with MS risk.11, 13-18 However, studies on the relationship between smoking and

MS progression show mixed results. Specifically, some studies demonstrate an effect of smoking on the conversion from a first episode of CNS demyelination—clinically isolated syndrome (CIS)—to clinically definite multiple sclerosis (CDMS),19, 20 as well as on the conversion from a relapsing-remitting multiple sclerosis (RRMS) course to a secondary-progressive multiple sclerosis (SPMS) course.21, 22 Other studies show no effect.23-25 There have been mixed results from studies investigating the relationship between different forms of second-hand smoke exposure and MS. While some studies demonstrate an increased MS risk in relation to passive26, 27 and maternal prenatal smoke exposure,28 others do not.29, 30

Previous systematic reviews with meta-analyses examining the association

57 between smoking and MS risk have demonstrated a statistically significant relationship with an effect estimate of approximately 1.5.31-35 The two most recent reviews examined the effect of passive smoking on MS, and while one showed a statistically significant relationship,35 the other did not.34 Only one previous review examined the association between smoking and MS progression, and did not show a statistically significant association between smoking and risk of SPMS.31

Previous reviews have several limitations and research gaps remain. Specifically, there is a need to include both more recently published and previously missed studies that provide direct and indirect evidence relevant to the smoking-MS risk association. Also, systematic review and meta-analysis examination of the smoking-MS progression association is minimal and non-current; studies investigating the progression from CIS to

CDMS (a critical stage in the disease process) in relation to smoking have not yet been subjected to systematic review and meta-analysis. In addition, systematic reviews with meta-analyses have only recently begun to examine the relationship between passive smoking and MS. This creates an even greater need to collate and summarize all studies that investigate different forms of second-hand smoke exposure, which could represent important MS risk factors that require further examination. Finally, the Bradford Hill criteria for causation36 provide a useful tool consisting of nine criteria used to evaluate research evidence to assess whether a relationship between an exposure and a disease outcome is likely to be causal. Hill’s criteria include Strength, Consistency, Specificity,

Temporality, Biological Gradient, Plausibility, Coherence, Experimental Evidence, and

Analogy. However, Hill’s criteria have previously only been used in narrative reviews37, 38 evaluating the smoking-MS association. While Hill’s criteria have been used in

58 systematic reviews and meta-analyses assessing other relationships,39, 40 this has not been done for the smoking-MS relationship. Applying Hill’s criteria to data compiled through systematic review and meta-analytic methods would be a more comprehensive approach to evaluating the potential causality of the association.

Therefore, the primary objective of this research was to a) examine the association between smoking and both MS risk and MS progression (i.e., CIS to CDMS, and RRMS to SPMS) through a systematic review and meta-analysis and b) use Hill’s criteria to more comprehensively assess the causality of each association. A secondary objective was to briefly summarize the literature on the relationship between different forms of second-hand smoke exposure (i.e., passive, prenatal) and MS through a systematic review.

2. Methods

A review protocol is available through PROSPERO: International Prospective

Register of Systematic Reviews and can be retrieved at http://www.crd.york.ac.uk/PROSPERO/. (Registration number: CRD42015025039)

2.1 Search Strategy

The following databases were searched for published studies up until July 28,

2015: Medline via PubMed (1965-current), EMBASE: Excerpta Medica & EMBASE

Classic via Ovid (1947-current), CINAHL Plus with Full Text via EBSCO (1937- current), PsycInfo via Ovid (1806-current), and Cochrane Library (1992-current). The search was restricted to English-language articles and used the following phrase:

59 (“multiple sclerosis” OR “disseminated sclerosis” OR “encephalomyelitis disseminata”

OR “clinically isolated syndrome” OR “clinically isolated syndromes” OR demyelinat*)

AND (smok* OR cigar* OR pipe OR pipes OR tobacco* OR nicotine* OR cotinine*).

No publication period limits were applied to the search. Reference lists of relevant studies and review articles were scanned for additional studies.

2.2 Eligibility Criteria

The review included peer-reviewed original research studies on humans with a case-control, cohort, or cross-sectional design that examined the relationship between smoking and MS. Specifically, studies investigating MS risk or progression (i.e., CIS to

CDMS, and RRMS to SPMS) for active, passive, or prenatal smoking versus non- smoking, using participants with a confirmed or self-reported MS diagnosis were included. Studies were also included in the review if they examined the interaction between smoking and other MS risk factors; this formed part of the evaluation of causality. For case-control studies to be included in the review, an appropriate comparator group must have been utilized, such as a group that is healthy or drawn from a general population, to minimize the risk of particular diseases or conditions confounding the relationship of interest.

2.3 Study Selection

All studies retrieved from the search were exported from the databases and imported to EndNote X3.0.1 (Thomson Reuters, New York City, NY, USA), where duplicate studies were located and removed. One reviewer (M.D.) screened the studies

60 using a piloted list of questions (e.g., Does the study investigate MS? Is the study an original research study? Appendix A) to ensure that studies complied with the eligibility criteria. First, the titles of all studies were screened, followed by the abstracts of those studies not excluded based on their respective titles. A second reviewer (K.H.) screened a random selection of 10% of the same titles and abstracts to verify inclusion and exclusion decisions, and establish an appropriate level of inter-rater reliability. The level of inter- rater reliability was determined as the percentage of titles and abstracts on which the reviewers agreed (the number of agreements divided by the total number of titles and abstracts screened by the two reviewers). The full-text publications of potentially relevant studies whose abstracts were not excluded were retrieved and assessed for inclusion by

M.D., and any uncertainties were discussed between M.D. and K.H. to reach a consensus.

Opinions provided by both reviewers were given equal weight during all discussions of uncertainties.

2.4 Data Extraction

Data were extracted from the included studies using a piloted Microsoft Excel spreadsheet. Data obtained from the eligible studies and stored in the spreadsheet included: First author and year of publication, study objective and design, country of origin, sample size (initial number of participants and final number with outcome of interest, where applicable), age, sex (female to male ratio), source and disease history of study groups, smoking status and method for ascertaining smoking, MS course (initial

MS types or sequence of progression if provided and applicable), method of MS diagnosis, confounders controlled (matched or adjusted), and reported odds ratio (OR),

61 risk ratio (RR), or hazard ratio (HR) with 95% confidence interval (CI). If a study provided an effect estimate, either alone or alongside raw data (e.g., an OR representing the odds of developing MS among smokers compared to the odds of developing MS among non-smokers), this value was used in the analysis (Appendix B). However, if effect estimates were not provided, either raw data (if available from the studies) were used to calculate an effect estimate and 95% CI (i.e., total number of cases and controls, and number of smokers in each participant group), or results were qualitatively reported.

Analyses of dose-response relationships (statistically tested or observed) and tests for interactions between smoking and other MS risk factors were considered and results were noted. Consideration was also given to investigations and discussions of possible mechanisms underlying the relationship between smoking and MS, as well as associations analogous to the relationship of interest.

2.5 Quality Assessment

The Grading of Recommendations Assessment, Development, and Evaluation

(GRADE) approach was used by one reviewer (M.D.) to evaluate the quality of evidence overall for each outcome of interest, namely MS risk, and the conversion from CIS to

CDMS, and RRMS to SPMS. The second reviewer (K.H.) was consulted in cases where uncertainty existed around whether grade modifications were warranted. Only studies investigating active smoking were assessed. A detailed description of the GRADE framework has been published.41 Briefly, the quality of evidence for the three outcomes was assigned a grade of high, moderate, low, or very low depending on the following factors outlined by GRADE: Study design, study limitations, inconsistency of results,

62 indirectness of evidence, imprecision of results, publication bias, a large effect estimate, presence of a dose-response relationship, and residual confounding. Specifically, observational studies were initially assigned a low quality of evidence grade. The grade was reduced if there were limitations in design and execution, inconsistent effect estimates, indirect evidence, imprecise effect estimates (and therefore wider CIs), or publication bias. On the other hand, the grade was increased if there was a large or very large effect estimate, dose-response relationship, or an effect of residual confounding

(further details in Appendix C).

2.6 Meta-Analysis

A random-effects meta-analysis was conducted using the generic inverse variance

(IV) method to calculate the pooled effect estimates (OR/RR or HR) and 95% CIs for the effect of active smoking on both MS risk and progression. Greater weight (software generated) was typically given to studies whose standard errors (SEs) were smaller, which tended to be larger studies, with the aim of achieving more precise pooled estimates.42 Previously reported formulas43 were utilized to obtain the data necessary to use the generic IV method, including converting the individual effect estimates and their corresponding errors into natural logarithms for input into the software program, which would then convert them back into ratio measurements (OR/RR or HR) and 95% CIs for display. A random-effects meta-analysis is a desirable approach for the current review, which assumes that the effect estimates across individual studies differ.44 Such heterogeneity is found to be a common occurrence, but using a random-effects approach

63 allows for heterogeneity to be integrated in an analysis, resulting in a larger amount of variation in the overall effect estimate.44

Separate analyses were conducted for each of the three outcomes of interest, with forest plots created. As in previous reviews,31, 35 for MS risk, both a conservative (only studies in which smoking was considered before MS onset) and non-conservative (all studies) analysis were conducted. The conservative analysis establishes temporality from smoking to MS risk, which is required to infer causality.45 The non-conservative analysis allows for the inclusion of additional evidence which, given similar results, strengthens the overall conclusions drawn. A subgroup analysis was also carried out to compare the risk estimates for different study designs, smoking statuses, MS diagnostic methods, and methods for ascertaining self-reported smoking, in line with categories or variations investigated in previous reviews.31, 33, 35

Given that MS is a rare condition, the rare disease assumption46 was used to combine ORs and RRs from the individual studies. Under this assumption, the OR approaches the RR when the prevalence of a disease is low. If the studies did not provide effect estimates and 95% CIs to be used in the analysis, these were calculated based on the reported number of smoking events in each participant group if provided. Where possible, estimates for current smoking and past smoking provided by a study were combined to obtain a measure of ever-smoking to be used in the appropriate analysis (see

Appendix B for individual provided and calculated effect estimates for smoking-MS risk association). The I2 statistic47, 48 was used to detect and interpret heterogeneity among the studies included in each analysis. Specifically, the value of I2 determined whether heterogeneity was possibly unimportant (0%-40%), moderate (30%-60%), substantial

64 (50%-90%), or considerable (75%-100%).49 The possible presence of publication bias was assessed using funnel plots,50 which graph the individual study effect estimates against their SEs, whereby a relatively symmetrical inverted funnel shape indicates a lower likelihood of bias. Funnel plots were not provided for analyses of <10 studies as the plots could be misleading due to difficulty in assessing symmetry.51 If applicable, analyses were repeated with and without outliers.

The meta-analysis was performed using Review Manager 5.3 (Cochrane

Collaboration, Copenhagen, DK).

2.7 Application of Hill’s Criteria

The Bradford Hill criteria for causation36 were applied to data from the systematic review and meta-analysis to evaluate the causal nature of the relationship between active smoking and each outcome (MS risk; CIS to CDMS; RRMS to SPMS). As Hill’s criteria are originally stated very generally, the criteria were further defined for purposes of the present review by adding specific parameters to be used to assess whether or not each criterion was satisfied:

1. Strength: The pooled effect estimate for each relationship was interpreted as a

statistically significant strong (3.0-8.0), moderate (1.8-3.0), modest (1.4-1.7), or

weak (1.1-1.3) relationship52 in the expected direction; a statistically non-

significant effect estimate was interpreted as no relationship.

2. Consistency: ≥50% vs. <50% of all included studies for each association showed

an effect of similar strength in the expected direction.

65 3. Temporality: ≥50% vs. <50% of all included studies for each association

established a temporal direction by design.

4. Biological Gradient: ≥50% vs. <50% of tests for a dose-response relationship

either showed a statistically significant trend or an apparent change in the

magnitude of effect in the expected direction.

5. Plausibility: Possible mechanisms underlying each relationship, and statistically

significant interactions between smoking and other MS risk factors existed vs. did

not exist.

6. Coherence: The associations did not conflict with existing MS knowledge vs.

conflict observed.

7. Analogy: Analogous relationships existed vs. did not exist.

In a scoring system created for this review, one point was given for each criterion satisfied; criteria not satisfied were given zero points. The number of criteria satisfied were summed together to produce an overall causation score for each association ranging from 0-7. A causation score of 7 or 6, 5 or 4, and ≤3 represented strong, moderate, and weak evidence, respectively, supporting a cause-and-effect association. A similar evaluation process using different scoring systems has been previously employed by other systematic reviews and meta-analyses assessing other relationships.39, 40

From the original nine criteria, the Specificity criterion was deemed non- applicable to the current review given that there are known relationships between smoking and other diseases besides MS.53, 54 Furthermore, the Experimental Evidence criterion was non-applicable since studying smoking as a cause of disease under controlled conditions is unethical.

66 3. Results

3.1 Search Results

The search strategy initially produced 1870 citations. A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram55 illustrating the study selection process is presented in Figure 1. In terms of the level of inter-rater reliability achieved, the two reviewers had 90% agreement on the selection of titles and abstracts. After discussion of selection disagreements, full consensus was reached.

Following the exclusion of irrelevant titles, abstracts, and full-text articles, 56 studies were left to be included in the review (publication dates 1965-2015). Specifically, the studies were included in the relevant quantitative and/or qualitative synthesis, depending on the area of investigation in each study. The following meta-analyses were included in the current review: 1) Smoking vs. MS risk; 2) Smoking vs. CIS to CDMS; 3) Smoking vs. RRMS to SPMS. The following qualitative analyses were included: 1) Smoking vs.

CIS to CDMS; 2) Smoking vs. RRMS to SPMS; 3) Passive and prenatal smoke exposure vs. MS risk; and 4) Risk factor interactions. Appendix D provides a list of study references included under each analysis. Studies listed in the forest plots are not referenced in this manuscript if not discussed in the manuscript, but are listed with the other included studies in Appendix D. Characteristics of the studies included in each meta-analysis, such as the study design, sample sizes, and details pertaining to the population groups, exposure and outcomes of interest, are provided in Appendix E. There were 66 full-text articles excluded from the review, with reasons for exclusion listed in

Figure 1. Appendix F provides a list of excluded full-text study references and corresponding reasons for their exclusion.

67 Records(identified(through(database( searching((n(=(1870)( CINAHL(Plus(with(Full(Text:(129( Cochrane(Library:(13( EMBASE((Ovid):(1183( ! Medline((PubMed):(405( PsycInfo((Ovid):(140( (

Records(after(duplicates(within(databases( removed((n=1812)( CINAHL(Plus(with(Full(Text:(127( Cochrane(Library:(13( EMBASE((Ovid):(1130( Medline((PubMed):(403( PsycInfo((Ovid):(139( ( ( s( ( Records(after(duplicates(across(databases( removed((n=1269)( CINAHL(Plus(with(Full(Text:(50( Cochrane(Library:(0( EMBASE((Ovid):(787( Medline((PubMed):(403( PsycInfo((Ovid):(29( ( Titles(excluded((n(=(811)( ( CINAHL(Plus(with(Full(Text:(38( ( Titles(screened((n(=(1269)( Cochrane(Library:(0( ( EMBASE((Ovid):(555( Medline((PubMed):(193( PsycInfo((Ovid):(25( (

Abstracts(screened((n(=(458)( Abstracts(excluded((n(=(338)( CINAHL(Plus(with(Full(Text:(12( Cochrane(Library:(0( EMBASE((Ovid):(223( Medline((PubMed):(100( PsycInfo((Ovid):(3( ( Additional(records(identified(through( FullUtext(articles(assessed(for( reference(list(scanning((n(=(2)( eligibility((n(=(122)( FullUtext(articles(excluded((n(=(66)( ( Smoking(not(investigated:(2( ( Smoking(not(analyzed(as(independent( Studies(included(in(review((n(=(56)( variable:(3( ( ( Effects(of(smoking(on(MS(not(reported:(10( Smoking(&(MS(Risk((MetaUAnalysis):(36( ( ( Impossible(to(isolate(effects(of(smoking:(2( Smoking(&(CIS(to(CDMS((MetaUAnalysis):(3( ( ( Participants(matched(on(smoking:(1( Smoking(&(RRMS(to(SPMS((MetaUAnalysis):(5( ( ( Outcomes(of(interest(not(assessed:(25( Smoking(&(CIS(to(CDMS((Qualitative):(2( ( ( Diagnosed(MS(not(investigated:(2( Smoking(&(RRMS(to(SPMS((Qualitative):(2( ( ( NonUoriginal(research(study:(2( Passive(Smoking(&(MS(Risk((Qualitative):(8( ( ( Ecological(study:(4( Prenatal(Smoking(&(MS(Risk((Qualitative):(4( ( ( No(comparator(group:(3( Risk(Factor(Interactions((Qualitative):(7( ( ( Unhealthy(control(group:(4( **Note:(The(sum(of(above(studies(exceeds( (

Duplicate(data:(7( 56(as(some(were(used(in(more(than(one( ( analysis( Comparison(with(prior(published(data:(1( ( Figure 1. Flow diagram of study selection process depicting number of studies identified, excluded (including titles, abstracts, full-text articles), and included in review CIS = clinically isolated syndrome; CDMS = clinically definite multiple sclerosis; RRMS = relapsing- remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis

68 3.2 Smoking and MS Risk

The meta-analysis of the association between smoking and MS risk included 36 studies. Case-control and cross-sectional studies contributed 21,996 cases and 9,199,028 controls; another 764 cases developed in a population of 562,854 individuals contributed by cohort studies.

The conservative analysis included 23 studies in which smoking was considered before MS, and showed a statistically significant association between smoking and MS risk (OR/RR 1.54, 95% CI [1.46-1.63]; Figure 2) with statistically non-significant heterogeneity, χ2(22)=21.20, p=0.51; I2 = 0%. Of the total weight, the large study by

Hedstrom et al., 201314 took 32.9%, likely due to its narrow 95% CI of 1.36-1.65. The non-conservative analysis including all 36 studies produced similar results (OR/RR 1.46,

95% CI [1.32-1.61]; Figure 3), with the exception of demonstrating substantial heterogeneity, χ2(35)=137.20, p<0.00001; I2 = 74%. Compared to the conservative analysis, the total weight was distributed more evenly across the individual studies included in the non-conservative analysis, likely due to the inclusion of more studies. The funnel plots for the conservative (Figure 4) and non-conservative (Figure 5) analyses were generally symmetrical, indicating that publication bias was unlikely.

Figure 2. Meta-analysis and forest plot: Relationship between smoking and MS risk (conservative*) IV = inverse variance; CI = confidence interval Studies identified by first author and year. *Includes 23 studies where smoking considered before MS onset

Figure 3. Meta-analysis and forest plot: Relationship between smoking and MS risk (non-conservative*) IV = inverse variance; CI = confidence interval Studies identified by first author and year. *Includes all 36 studies

Figure 4. Funnel plot: Conservative* analysis, association of smoking with MS risk OR = odds ratio; SE = standard error *Includes 23 studies where smoking considered before MS onset

Figure 5. Funnel plot: Non-conservative* analysis, association of smoking with MS risk OR = odds ratio; SE = standard error *Includes all 36 studies

73 The subgroup analyses presented in Table 1 showed no statistically significant differences among study designs, MS diagnostic methods, and methods for ascertaining self-reported smoking; however, the risk of MS was greater among studies comparing ever-smokers with never-smokers than among those comparing current smokers with past and never-smokers in combination (p=0.02). There was one outlier56 identified among the studies included in the non-conservative analysis whose effect estimate was especially distant from the other estimates. Repeating the non-conservative and subgroup analyses without this outlier did not change the overall results.

Ten studies included in the meta-analysis presented a dose-response relationship between smoking and MS risk either in the form of a statistically significant trend or change in effect estimates in the expected direction;13-15, 57-63 however, three additional studies that tested for a dose-response relationship were not able to demonstrate it.12, 17, 64

Based on the GRADE approach, the overall quality of evidence for MS risk was low (Table 2). While this assessment could have been upgraded from low to moderate quality due to the presence of a dose-response relationship, there were serious study limitations that prevented upgrading. One example is the inclusion of studies in which the large majority had a retrospective case-control design that is prone to recall bias, a limitation highlighted in several of the studies themselves.30, 58, 65 Other examples of study limitations include the considerable use of smoking data based on self-report,64, 66, 67 as well as either a low response/participation rate or a noticeably different rate between cases and controls as demonstrated by some studies.14, 58, 66

74 Table 1. Smoking-MS risk effect estimates with 95% CIs for study subgroups

Subgroup Study OR/RR 95% CI p value* Count Case-control 28 1.50 1.38-1.64 Cohort 4 1.52 1.27-1.81 0.93

Cohort 4 1.52 1.27-1.81 Case-control/cross-sectional 32 1.45 1.30-1.62 0.69

Ever vs. never smoking 28 1.57 1.46-1.69 Current vs. past and never smoking 8 1.11 0.83-1.49 0.02

Smoking ascertainment using medical 20 1.42 1.28-1.57 records/interview Smoking ascertainment using questionnaire 15 1.50 1.27-1.78 0.56

MS diagnostic criteria 26 1.51 1.39-1.65 Other diagnostic methods 7 1.30 0.87-1.94 0.47

OR = odds ratio; RR = risk ratio; CI = confidence interval * χ2 test for differences between subgroups. p<0.05 statistically significant difference between subgroups

75 Table 2. Quality of evidence evaluation for MS risk and MS progression outcomes using the GRADE approach

GRADE FACTORS Study Publication Factors That OUTCOME Limitations* Inconsistency Indirectness Imprecision Bias Increase Quality QUALITY MS risk Serious No serious Direct No serious Unlikely Dose-Response ++, low limitations unexplained imprecision Relationship (+1) (-1)† inconsistency CIS!CDMS Serious Serious Direct Imprecision (-1)|| Inconclusive None +, very low limitations unexplained (-1)‡ inconsistency (-1) RRMS!SPMS Serious Serious Direct Imprecision (-1)¶ Inconclusive Dose-Response +, very low limitations unexplained Relationship (+1) (-1)§ inconsistency (-1) ! GRADE = Grading of Recommendations Assessment, Development, and Evaluation; CIS = clinically isolated syndrome; CDMS = clinically definite multiple sclerosis; RRMS = relapsing-remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis; CI = confidence interval

*All outcomes: Each outcome contained at least one retrospective study (e.g., case-control, retrospective cohort), most studies relied on self-reported smoking data, many studies did not specify any blinding procedures used in the exposure or outcome assessments (where appropriate).

†Limitations include those that apply to all outcomes, and the following that apply to this outcome only: Self-reported MS diagnosis in one study,73 either response/participation rate < 80% or rate notably different between cases and controls in nine studies,11, 12, 14, 30, 58, 60, 66, 67, 73 > 20% lost to follow-up in one study,62 type of smoking ascertainment not stated in one study,59 MS diagnostic method not stated in three studies.57, 59, 74

‡Limitations include those that apply to all outcomes, and the following that apply to this outcome only: Source of cohort not stated in one study,19 MS diagnostic method not stated in one study.75

§Limitations include those that apply to all outcomes, and the following that apply to this outcome only: Source of cohort not stated in one study,19 response/participation rate < 80% in two studies,22, 24 > 20% lost to follow-up in one study.24

||95% CI in meta-analysis included no effect and notable benefit and harm.

¶While not including no effect, 95% CI in meta-analysis conveyed borderline significance up to notable harm; many individual studies produced wide CIs.

76 3.2.1 Risk Factor Interactions

In addition to the 36 smoking-MS risk studies, four studies specifically investigated interactions between MS risk factors, and some of the 36 studies conducted further analyses on interactions pertaining to MS risk. While there were studies that demonstrated no statistically significant interactions between smoking and both EBV68-70 and HLA alleles,60, 69, 70 others showed that these risk factors interacted with smoking.60, 71

Recent studies also showed significant interactions between smoking and both non- consumption of alcohol72 and non-HLA alleles.67

3.3 Smoking and Conversion from CIS to CDMS

The meta-analysis of the effect of smoking on the conversion from CIS to CDMS included three studies comprised of 755 CDMS patients developed from 1261 CIS patients. The analysis showed a statistically non-significant association (HR 1.13, 95% CI

[0.73-1.76]; Figure 6) with substantial heterogeneity, χ2(2)=7.68, p=0.02; I2 = 74%.

Two prospective cohort studies were not included in the meta-analysis due to insufficient data to calculate the unreported HR or other appropriate statistics,75 or calculation of the required statistics from provided data19 not performed for this meta- analysis due to statistical methods previously cited as likely being less reliable than other methods.76 One of these studies using self-reported smoking data from 125 CIS patients showed that smoking increased the risk of progression from CIS to CDMS.19 However, the second study of 194 CIS patients used cotinine as a measure of smoke exposure and showed that smoking had no effect.75

Figure 6. Meta-analysis and forest plot: Relationship between smoking and conversion from CIS to CDMS CIS = clinically isolated syndrome; CDMS = clinically definite multiple sclerosis; IV = inverse variance; CI = confidence interval Studies identified by first author and year.

78 Only one study investigated a possible dose-response effect for the conversion from CIS to CDMS, and found none.23 Due to individual study limitations, including important study characteristics not stated in some studies,19, 75 as well as inconsistent results across studies and imprecise meta-analysis results, the quality of evidence for the progression from CIS to CDMS was downgraded from low to very low (Table 2).

3.4 Smoking and Conversion from RRMS to SPMS

The meta-analysis of the association of smoking with SPMS risk included five studies comprised of 379 SPMS patients developed from 1837 RRMS patients. The analysis yielded an association just achieving statistical significance (HR 1.80, 95% CI

[1.04-3.10]; Figure 7) with substantial heterogeneity, χ2(4)=12.68, p=0.01; I2 = 68%.

Two prospective cohort studies were not included in this meta-analysis. As previously noted in one case, potentially unreliable methods for calculating the required statistics (e.g., HR) were avoided.19 The second study for this outcome measured the association between smoking and the conversion from RRMS to SPMS per pack year smoked77 rather than comparing smoking and non-smoking as in all other studies.

Nonetheless, both studies demonstrated a statistically significant effect of smoking on the conversion from RRMS to SPMS by either comparing smokers and non-smokers in 203

RRMS patients,19 or per pack-year smoked in 148 RRMS patients.77

Figure 7. Meta-analysis and forest plot: Relationship between smoking and conversion from RRMS to SPMS RRMS = relapsing-remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis; IV = inverse variance; CI = confidence interval Studies identified by first author and year.

80 While one study did not demonstrate a dose-response relationship between smoking and SPMS risk,24 three studies showed evidence of a dose-response relationship.22, 77, 78 However, the quality of evidence for the association between smoking and SPMS risk was downgraded from low to very low given the presence of study limitations, as well as inconsistent and imprecise results (Table 2). Some examples of study limitations include the use of self-reported smoking data22, 78 and a low response/participation rate as shown by some studies.22, 24

3.5 Assessment of Causality Using Hill’s Criteria

An evaluation of the causal nature of the relationship between active smoking and both MS risk and MS progression outcomes using Hill’s criteria is displayed in Table 3.

Based on data from the systematic review and meta-analysis, and the scoring protocol developed for this review, there is strong evidence supporting a causal association between smoking and MS risk (causation score of 6 out of 7), while there is only moderate evidence of a causal role of smoking in the progression from CIS to CDMS

(4/7) and RRMS to SPMS (5/7).

81 Table 3. Causality evaluation of the association of smoking with MS risk and MS progression outcomes using Hill’s criteria

Hill’s Criteria MS Risk CISàCDMS RRMSàSPMS Strength 0 0 0 Consistency 1 0 0 Temporality 1 1 1 Biological Gradient 1 0 1 Plausibility 1 1 1 Coherence 1 1 1 Analogy 1 1 1 Causation Score 6 4 5 Interpretation Strong evidence of Moderate evidence Moderate evidence causality of causality of causality

CIS = clinically isolated syndrome; CDMS = clinically definite multiple sclerosis; RRMS = relapsing- remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis

82 3.6 Passive/Prenatal Smoke Exposure and MS Risk

Given that passive and prenatal smoke exposure may play a potential role in MS risk, the relationship between second-hand smoke exposure and MS risk was briefly explored. The association between passive smoke exposure and MS risk was examined by eight studies, with mixed results. Four studies did not show a statistically significant relationship between passive smoking and MS, including three with a case-control design that investigated exposure primarily among adults,14, 30, 79 and one that specifically investigated childhood exposure in a cohort of female nurses.80 Among the four remaining case-control studies that showed a statistically significant relationship between passive smoking and MS, one examined exposure strictly among children,81 one used cotinine as a marker of smoke exposure,61 and two demonstrated a statistically significant dose-response relationship between duration of passive smoke exposure and MS.26, 27

Four studies examined the relationship between maternal prenatal smoke exposure and MS risk in offspring. One case-control study involving 35 hospitalized MS cases with smoking data demonstrated a strong association.28 The other three studies did not show a statistically significant relationship, including two larger case-control studies,29, 30 and a large prospective cohort study.80 One study examined a possible dose- response effect, but did not demonstrate that MS risk increased with cumulative number of cigarettes smoked by the mother.29

4. Discussion

This systematic review and meta-analysis examined the relationship between smoking and both MS risk and MS progression (i.e., CIS to CDMS, and RRMS to

83 SPMS), and subsequently used the Bradford Hill criteria to further evaluate causation.

Overall, a statistically significant association exists between smoking and both MS risk and the conversion from RRMS to SPMS. However, there is a statistically non-significant association between smoking and the progression from CIS to CDMS. Based on data from the review, as well as the application of Hill’s criteria, there is strong evidence of a causal relationship between smoking and MS risk, but only moderate evidence of a causal role of smoking in MS progression.

While the findings were generally consistent with those from the narrative reviews that also used Hill’s criteria,37, 38 the present review involved a more rigorous and objective methodology to review all relevant studies and apply Hill’s criteria to the most reliable study findings to evaluate causality. Overall, the large majority of studies included for each of the three outcomes firmly established a temporal direction from smoking to the outcome of interest, which is necessary to infer causation.45 In this review, temporality was established in many studies either through the use of a prospective cohort design, or smoking was considered prior to MS onset in case-control and cross- sectional designs.

4.1 Smoking and MS Risk

Based on the meta-analysis, smoking led to a modest 50% increase in MS risk in comparison to non-smoking, which is in agreement with previous meta-analyses.31-35 The conservative and non-conservative analyses yielded similar results, both of which showed a statistically significant relationship between smoking and MS risk. In addition, the possibility of publication bias was unlikely given that the funnel plots for both analyses

84 were, for the most part, symmetrical, lending further weight to the results. Moreover, the majority of individual studies that tested for a dose-response relationship provided evidence of such a gradient in the expected direction, thus supporting a causal role of smoking in MS. In general, studies showed that MS risk increased with greater duration of smoking,14 higher levels of cotinine,61 cumulative numbers of cigarettes smoked daily,14, 57, 62, 63 cumulative numbers of packs smoked daily,59 and greater pack-years smoked13-15, 58, 60 as evidenced by a greater MS risk among heavier smokers.

There was substantial heterogeneity in the non-conservative analysis, meaning that most of the variation in the individual effect estimates was more likely the result of true differences as opposed to chance alone.47 Using a subgroup analysis to explore the possible sources of heterogeneity, it was found that different study designs, MS diagnostic methods, and methods for ascertaining self-reported smoking had no statistically significant effect on the association between smoking and MS risk. However, an approximately 2.5 times greater MS risk was observed among the larger group of studies that compared ever-smokers with never-smokers than among the studies that compared current smokers with the combination of past and never-smokers. Classifying a past smoker as a non-smoker may lead to a weaker association between smoking and MS given that those who have quit smoking are still at a higher risk of MS in comparison to non-smokers.34, 35 It is also likely that some heterogeneity resulted from the one outlying study included in the non-conservative analysis,56 which had no major effect on the overall results. Despite being cross-sectional, its results suggested a statistically significant protective effect of smoking on MS risk. On the other hand, the findings showed that heterogeneity was likely unimportant in the conservative analysis, whereby

85 virtually all included studies compared ever-smokers with never-smokers, and the outlier had been removed. Overall, the estimates in the conservative analysis were consistent in that a modest effect of smoking on MS risk was evident across the majority of studies, and all studies demonstrated an effect in the expected direction. This consistency provides additional evidence of a causal role of smoking in MS risk.

4.2 Smoking and MS Progression

Regarding MS progression, the meta-analysis showed that smoking contributed to a moderate 80% increase in SPMS risk compared to non-smoking, which is in close agreement with a previous meta-analysis.31 However, while the relationship in the previous review was statistically non-significant, the relationship in the current review was statistically significant, likely due to the inclusion of a more recent study showing a statistically significant relationship between smoking and SPMS risk.22 The pooled effect estimate in the current review was corroborated by the results from the two additional studies not included in the meta-analysis that also demonstrated a statistically significant association between smoking and the progression from RRMS to SPMS.19, 77 It was also supported by evidence of a dose-response relationship between secondary progression and the number of cigarettes smoked daily22 and pack-years smoked77, 78 as evidenced by tests for such a relationship, providing further evidence of causality.

On the other hand, the meta-analysis showed no statistically significant relationship between smoking and the conversion from CIS to CDMS, and no individual studies provided evidence of a dose-response gradient. Mixed results were evident in the two studies not included in the meta-analysis as one study showed a statistically

86 significant effect of smoking,19 while the second study did not,75 further confirming that insufficient evidence exists to support a causal role of smoking in the progression from

CIS to CDMS.

While the MS progression outcomes exist along the same disease continuum, their results differed in this review. Since MS presents more neurological damage (i.e., characterized by multiple regions of CNS demyelination) compared to CIS (i.e., characterized by a first episode of CNS demyelination),82 it is possible that the greater damage could make RRMS patients more susceptible to the harms of cigarette smoke compared to CIS patients. Alternatively, results may have been impacted by the use of different study methods between the two outcomes, such as the smoking status under study (e.g., predominantly current smokers for CIS to CDMS, and ever-smokers for

RRMS to SPMS), methods for ascertaining smoking (e.g., self-report for both outcomes, but measures of cotinine included for CIS to CDMS as well), and confounders controlled

(e.g., CIS symptoms and lesions for CIS to CDMS, and disease duration and relapses for

RRMS to SPMS).

Overall, the meta-analysis for both progression outcomes was based on a small number of studies, and as such, funnel plots were not generated given their potential to be misleading in such circumstances.51 As such, there is a possibility that publication bias exists, meaning that published studies in this area, especially in the area of smoking and

SPMS risk, may have been more likely to show an effect, while studies not showing an effect may have failed to be published.50 Furthermore, the meta-analysis for both outcomes yielded substantial heterogeneity across studies, indicating that most of the differences in the effect estimates were not likely the result of chance alone.47 Instead,

87 they may have been due to variation in the setting, study design, population groups, and type of smoke exposure measures used in the individual studies. As a result, the findings were generally inconsistent across the studies included for each outcome. The strength of the association between smoking and SPMS risk varied considerably from weak, to moderate, to very strong. Similarly, alternate directions of an effect and levels of statistical significance were observed among the individual estimates and statistical tests for the association between smoking and the progression from CIS to CDMS. Overall, these factors make it difficult to reach a definite conclusion for the two progression outcomes regarding the causal role of smoking.

4.3 Mechanisms, Interactions, Analogies Relevant to the Smoking-MS Association

The relationship between smoking and both MS risk and MS progression is plausible and coherent given that there are possible biological mechanisms underlying each relationship, and the relationships are consistent with what is already known about

MS. This provides some evidence of a causal role of smoking towards these outcomes.

Mechanistically, it has been shown that nicotine increases the permeability of the blood- brain barrier,83 which may represent an early step in MS.84 Other chemical components of cigarette smoke may also be involved. For instance, it has been found that cyanide contributes to CNS demyelination,85 and nitric oxide has been linked to degeneration and conduction blockage of axons;86 axonal loss is representative of MS progression.87 There is also evidence of statistically significant interactions between smoking and known MS risk factors, including HLA alleles and EBV. Specifically, smoking may elicit autoimmunity in those who are genetically susceptible.88 However, the interaction

88 between smoking and HLA may not always occur, depending on the presence of other genetic factors. For instance, while an interaction was found between smoking and the

HLA-DRB1*15 allele, it was only observed among those without the more protective

HLA-A*02 allele.71 Regarding EBV, both smoking and EBV infection have been shown to stimulate cellular activity that may play a critical role in MS development.60 However, the interaction between smoking and EBV may be affected by age given the evidence of a negative interaction among younger subjects68 and a positive interaction among older subjects,60, 68 again suggesting that an interaction may only occur under certain circumstances.

Relationships analogous to the association between smoking and both MS risk and MS progression exist, as shown by research demonstrating a relationship between smoking and other autoimmune diseases. Specifically, associations have been found between smoking and rheumatoid arthritis, Crohn’s disease, and ulcerative colitis risk,89 as well as increased risk of systemic lupus erythematosus90 and Graves’ disease.91 It was also found that smoking is associated with disease severity in several autoimmune disorders.91, 92 These analogies provide further evidence to support causality.

4.4 Passive/Prenatal Smoke Exposure and MS Risk

The relationship between both passive and prenatal smoke exposure and MS risk was briefly explored to draw attention to other factors that could also play an important role in MS. Regarding passive smoke exposure, results were mixed and heterogeneity existed among the studies in terms of age groups studied and types of smoking data used.

Likewise, there were mixed findings from two systematic reviews and meta-analyses

89 regarding whether or not a statistically significant relationship existed.34, 35 This area of research would benefit from additional studies and a subsequent subgroup analysis to determine if the relationship between passive smoking and MS only exists in certain circumstances. When it comes to maternal prenatal smoke exposure, the evidence thus far points to no association with MS risk. However, additional studies need to be conducted with a larger and more representative sample of MS cases prior to conducting an informative meta-analysis.

Strengths and limitations. The current review used a comprehensive search strategy, and included several databases, reducing the likelihood of missing important studies. Another strength of the review was the inclusion of a thorough quality assessment (GRADE) of included studies, allowing for a more meticulous evaluation of study evidence. Furthermore, distinguishing itself from previous systematic reviews and meta-analyses, the current review used Hill’s criteria to evaluate causation. This evaluation, in turn, allowed for a more complete examination of important aspects of the association between smoking and both MS risk and progression based on all relevant studies.

While the association between smoking and MS risk was based on many studies that included a large sample of participants, the number of studies included in the analysis of MS progression was comparatively small. As such, the results pertaining to

MS progression should be taken with caution. In addition, the quality of study evidence for each outcome was either low or very low. However, this can be largely attributed to the observational nature of studies examining the health effects of smoking, which is unavoidable given that such an investigation under controlled conditions is unethical (i.e.,

90 one cannot force participants to smoke in order to study its effects). As such, an initial quality grade of low is unavoidable. Furthermore, it may be challenging to improve this initial grade, especially given a GRADE factor such as study limitations, which can be difficult for study authors to avoid. One such limitation was the use of self-reported smoking data in most individual studies included in the review. However, while it has been shown that self-reported smoking habits tend to be underreported, such data is primarily relied upon in tobacco surveillance.93 Given such circumstances, the low and very low quality of evidence grade determined by the review should not be interpreted to mean that evidence was derived from poorly conducted studies. In fact, rather than being solely based on the methodological execution of individual studies, the grade was largely based on the collective examination of studies to evaluate a group of highly diverse factors (e.g., the inconsistency of effect estimates across studies, imprecision of effect estimates conveyed by the 95% CI in the meta-analysis, publication bias detected using funnel plots, dose-response relationship based on its frequency in studies). As such, despite the lower grades, the evidence obtained from the present review is essential in determining the causal role of smoking in MS.

Conclusions and implications. There is strong evidence of a causal role of smoking in MS risk. However, this review suggests only moderate evidence of a causal relationship between smoking and MS progression, namely from CIS to CDMS and

RRMS to SPMS. Overall, this review supports the addition of MS to the list of the more commonly known health outcomes related to smoking, such as the development and advancement of heart and lung diseases. As such, anti-smoking government policies and

91 smoking cessation programs should underscore the risk of developing MS and other neurological disorders as a further means of discouraging the public to smoke.

92 5. References

1. Multiple Sclerosis International Federation. Atlas of MS 2013. London, UK: Multiple

Sclerosis International Federation, 2013.

2. Murray TJ. Multiple sclerosis: The history of a disease. New York, NY: Demos

Medical Publishing, 2005.

3. World Health Organization. Atlas: Multiple sclerosis resources in the world. Geneva,

CH: WHO Press, 2008.

4. Jankosky C, Deussing E, Gibson RL, Haverkos HW. Viruses and vitamin D in the etiology of type 1 diabetes mellitus and multiple sclerosis. Virus Res 2012;163(2):424-30.

5. Abolfazli R, Samadzadeh S, Sabokbar T, et al. Relationship between HLA-DRB1*

11/15 genotype and susceptibility to multiple sclerosis in Iran. J Neurol Sci 2014;345(1-

2):92-6.

6. Sawcer S, Franklin RJ, Ban M. Multiple sclerosis genetics. Lancet Neurol

2014;13(7):700-9.

7. Almohmeed YH, Avenell A, Aucott L, Vickers MA. Systematic review and meta- analysis of the sero-epidemiological association between Epstein Barr virus and multiple sclerosis. PLoS One 2013;8(4):e61110.

8. Pakpoor J, Giovannoni G, Ramagopalan SV. Epstein-Barr virus and multiple sclerosis:

Association or causation? Expert Rev Neurother 2013;13(3):287-97.

9. Harandi AA, Pakdaman H, Sahraian MA. Vitamin D and multiple sclerosis. Iran J

Neurol 2014;13(1):1-6.

10. Pandit L, Ramagopalan SV, Malli C, D'Cunha A, Kunder R, Shetty R. Association of vitamin D and multiple sclerosis in India. Mult Scler 2013;19(12):1592-6.

93 11. O'Gorman C, Bukhari W, Todd A, Freeman S, Broadley SA. Smoking increases the risk of multiple sclerosis in Queensland, Australia. J Clin Neurosci 2014;21(10):1730-3.

12. Salzer J, Hallmans G, Nystrom M, Stenlund H, Wadell G, Sundstrom P. Smoking as a risk factor for multiple sclerosis. Mult Scler 2013;19(8):1022-7.

13. Asadollahi S, Fakhri M, Heidari K, Zandieh A, Vafaee R, Mansouri B. Cigarette smoking and associated risk of multiple sclerosis in the Iranian population. J Clin

Neurosci 2013;20(12):1747-50.

14. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Smoking and multiple sclerosis susceptibility. Eur J Epidemiol 2013;28(11):867-74.

15. Hernan MA, Olek MJ, Ascherio A. Cigarette smoking and incidence of multiple sclerosis. Am J Epidemiol 2001;154(1):69-74.

16. Hernan MA, Jick SS, Logroscino G, Olek MJ, Ascherio A, Jick H. Cigarette smoking and the progression of multiple sclerosis. Brain 2005;128(Pt 6):1461-5.

17. Maghzi AH, Etemadifar M, Heshmat-Ghahdarijani K, et al. Cigarette smoking and the risk of multiple sclerosis: A sibling case-control study in Isfahan, Iran.

Neuroepidemiology 2011;37(3-4):238-42.

18. Mansouri B, Asadollahi S, Heidari K, et al. Risk factors for increased multiple sclerosis susceptibility in the Iranian population. J Clin Neurosci 2014;21(12):2207-11.

19. Correale J, Farez MF. Smoking worsens multiple sclerosis prognosis: Two different pathways are involved. J Neuroimmunol 2015;281:23-34.

20. Di Pauli F, Reindl M, Ehling R, et al. Smoking is a risk factor for early conversion to clinically definite multiple sclerosis. Mult Scler 2008;14(8):1026-30.

94 21. Healy BC, Ali EN, Guttmann CR, et al. Smoking and disease progression in multiple sclerosis. Arch Neurol 2009;66(7):858-64.

22. Roudbari SA, Ansar MM, Yousefzad A. Smoking as a risk factor for development of secondary progressive multiple sclerosis: A study in Iran, Guilan. J Neurol Sci

2013;330(1-2):52-5.

23. Arikanoglu A, Shugaiv E, Tuzun E, Eraksoy M. Impact of cigarette smoking on conversion from clinically isolated syndrome to clinically definite multiple sclerosis. Int J

Neurosci 2013;123(7):476-9.

24. Koch M, van Harten A, Uyttenboogaart M, De Keyser J. Cigarette smoking and progression in multiple sclerosis. Neurology 2007;69(15):1515-20.

25. Kuhle J, Disanto G, Dobson R, et al. Conversion from clinically isolated syndrome to multiple sclerosis: A large multicentre study. Mult Scler 2015;21(8):1013-24.

26. Hedstrom AK, Bomfim IL, Barcellos LF, et al. Interaction between passive smoking and two HLA genes with regard to multiple sclerosis risk. Int J Epidemiol

2014;43(6):1791-8.

27. Hedstrom AK, Baarnhielm M, Olsson T, Alfredsson L. Exposure to environmental tobacco smoke is associated with increased risk for multiple sclerosis. Mult Scler

2011;17(7):788-93.

28. Mueller BA, Nelson JL, Newcomb PA. Intrauterine environment and multiple sclerosis: A population- based case-control study. Mult Scler 2013;19(1):106-11.

29. Montgomery SM, Bahmanyar S, Hillert J, Ekbom A, Olsson T. Maternal smoking during pregnancy and multiple sclerosis amongst offspring. Eur J Neurol

2008;15(12):1395-9.

95 30. Ramagopalan SV, Lee JD, Yee IM, et al. Association of smoking with risk of multiple sclerosis: A population-based study. J Neurol 2013;260(7):1778-81.

31. Handel AE, Williamson AJ, Disanto G, Dobson R, Giovannoni G, Ramagopalan SV.

Smoking and multiple sclerosis: An updated meta-analysis. PLoS One 2011;6(1):e16149.

32. Hawkes CH. Smoking is a risk factor for multiple sclerosis: A metanalysis. Mult

Scler 2007;13(5):610-5.

33. O'Gorman C, Broadley SA. Smoking and multiple sclerosis: Evidence for latitudinal and temporal variation. J Neurol 2014;261(9):1677-83.

34. Poorolajal J, Bahrami M, Karami M, Hooshmand E. Effect of smoking on multiple sclerosis: A meta-analysis. J Public Health (Oxf) 2016. doi:10.1093/pubmed/fdw030.

35. Zhang P, Wang R, Li Z, et al. The risk of smoking on multiple sclerosis: A meta- analysis based on 20,626 cases from case-control and cohort studies. PeerJ

2016;4:e1797.

36. Hill AB. The environment and disease: Association or causation? Proc R Soc Med

1965;58:295-300.

37. Loken-Amsrud KI, Lossius A, Torkildsen O, Holmoy T. Impact of the environment on multiple sclerosis. Tidsskr Nor Laegeforen 2015;135(9):856-60.

38. Wingerchuk DM. Smoking: Effects on multiple sclerosis susceptibility and disease progression. Ther Adv Neurol Disord 2012;5(1):13-22.

39. Henriksen M, Creaby MW, Lund H, Juhl C, Christensen R. Is there a causal link between knee loading and knee osteoarthritis progression? A systematic review and meta-analysis of cohort studies and randomised trials. BMJ Open 2014;4(7):e005368.

96 40. Mente A, de Koning L, Shannon HS, Anand SS. A systematic review of the evidence supporting a causal link between dietary factors and coronary heart disease. Arch Intern

Med 2009;169(7):659-69.

41. Guyatt GH, Oxman AD, Kunz R, Vist GE, Falck-Ytter Y, Schunemann HJ. What is

"quality of evidence" and why is it important to clinicians? BMJ 2008;336(7651):995-8.

42. Higgins JPT, Green S (editors). Section 9.4.3: A generic inverse-variance approach to meta-analysis. In: Cochrane handbook for systematic reviews of interventions (version

5.1.0) [updated March 2011]. Cochrane Collaboration, 2011. Available at: www.handbook.cochrane.org.

43. Higgins JPT, Green S (editors). Section 7.7.7.2: Obtaining standard errors from confidence intervals and P values: Absolute (difference) measures; Section 7.7.7.3:

Obtaining standard errors from confidence intervals and P values: Ratio measures. In:

Cochrane handbook for systematic reviews of interventions (version 5.1.0) [updated

March 2011]. Cochrane Collaboration, 2011. Available at: www.handbook.cochrane.org.

44. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials

1986;7(3):177-88.

45. Rothman KJ, Greenland S. Causation and causal inference in epidemiology. Am J

Public Health 2005;95 Suppl 1:S144-50.

46. Clayton D, Hills M. Statistical models in epidemiology. New York, NY: Oxford

University Press, 1993.

47. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta- analyses. BMJ 2003;327(7414):557-60.

97 48. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med

2002;21(11):1539-58.

49. Higgins JPT, Green S (editors). Section 9.5.2: Identifying and measuring heterogeneity. In: Cochrane handbook for systematic reviews of interventions (version

5.1.0) [updated March 2011]. Cochrane Collaboration, 2011. Available at: www.handbook.cochrane.org.

50. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315(7109):629-34.

51. Anzures-Cabrera J, Higgins JP. Graphical displays for meta-analysis: An overview with suggestions for practice. Res Synth Methods 2010;1(1):66-80.

52. Schoenbach VJ, Rosamund WD. Relating risk factors to health outcomes. In:

Understanding the fundamentals of epidemiology--An evolving text. Chapel Hill, NC:

Department of Epidemiology, School of Public Health, University of North Carolina,

2000:161-207.

53. Fagerstrom K. The epidemiology of smoking: Health consequences and benefits of cessation. Drugs 2002;62 Suppl 2:1-9.

54. U.S. Department of Health and Human Services. The health consequences of involuntary exposure to tobacco smoke: A report of the Surgeon General. Atlanta, GA:

Centers for Disease Control and Prevention, 2006.

55. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. PLoS Med 2009;6(7):e1000097.

98 56. Ploughman M, Beaulieu S, Harris C, et al. The Canadian survey of health, lifestyle and ageing with multiple sclerosis: Methodology and initial results. BMJ Open

2014;4(7):e005718.

57. Ghadirian P, Dadgostar B, Azani R, Maisonneuve P. A case-control study of the association between socio-demographic, lifestyle and medical history factors and multiple sclerosis. Can J Public Health 2001;92(4):281-5.

58. Hedstrom AK, Baarnhielm M, Olsson T, Alfredsson L. Tobacco smoking, but not

Swedish snuff use, increases the risk of multiple sclerosis. Neurology 2009;73(9):696-

701.

59. Mouhieddine TH, Darwish H, Fawaz L, Yamout B, Tamim H, Khoury SJ. Risk factors for multiple sclerosis and associations with anti-EBV antibody titers. Clin

Immunol 2015;158(1):59-66.

60. Simon KC, van der Mei IA, Munger KL, et al. Combined effects of smoking, anti-

EBNA antibodies, and HLA-DRB1*1501 on multiple sclerosis risk. Neurology

2010;74(17):1365-71.

61. Sundstrom P, Nystrom L, Hallmans G. Smoke exposure increases the risk for multiple sclerosis. Eur J Neurol 2008;15(6):579-83.

62. Thorogood M, Hannaford PC. The influence of oral contraceptives on the risk of multiple sclerosis. Br J Obstet Gynaecol 1998;105(12):1296-9.

63. Villard-Mackintosh L, Vessey MP. Oral contraceptives and reproductive factors in multiple sclerosis incidence. Contraception 1993;47(2):161-8.

64. Jafari N, Hoppenbrouwers IA, Hop WC, Breteler MM, Hintzen RQ. Cigarette smoking and risk of MS in multiplex families. Mult Scler 2009;15(11):1363-7.

99 65. Al-Afasy HH, Al-Obaidan MA, Al-Ansari YA, et al. Risk factors for multiple sclerosis in Kuwait: A population-based case-control study. Neuroepidemiology

2013;40(1):30-5.

66. Bjornevik K, Riise T, Cortese M, et al. Level of education and multiple sclerosis risk after adjustment for known risk factors: The EnvIMS study. Mult Scler 2016;22(1):104-

11.

67. Briggs FB, Acuna B, Shen L, et al. Smoking and risk of multiple sclerosis: Evidence of modification by NAT1 variants. Epidemiology 2014;25(4):605-14.

68. Salzer J, Stenlund H, Sundstrom P. The interaction between smoking and Epstein-

Barr virus as multiple sclerosis risk factors may depend on age. Mult Scler

2014;20(6):747-50.

69. Simon KC, Schmidt H, Loud S, Ascherio A. Risk factors for multiple sclerosis, neuromyelitis optica and transverse myelitis. Mult Scler 2015;21(6):703-9.

70. Sundqvist E, Sundstrom P, Linden M, et al. Lack of replication of interaction between

EBNA1 IgG and smoking in risk for multiple sclerosis. Neurology 2012;79(13):1363-8.

71. Hedstrom AK, Sundqvist E, Baarnhielm M, et al. Smoking and two human leukocyte antigen genes interact to increase the risk for multiple sclerosis. Brain 2011;134(Pt

3):653-64.

72. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Alcohol as a modifiable lifestyle factor affecting multiple sclerosis risk. JAMA Neurol 2014;71(3):300-5.

73. Riise T, Nortvedt MW, Ascherio A. Smoking is a risk factor for multiple sclerosis.

Neurology 2003;61(8):1122-4.

100 74. Kappus N, Weinstock-Guttman B, Hagemeier J, et al. Cardiovascular risk factors are associated with increased lesion burden and brain atrophy in multiple sclerosis. J Neurol

Neurosurg Psychiatry 2016;87(2):181-7.

75. Horakova D, Zivadinov R, Weinstock-Guttman B, et al. Environmental factors associated with disease progression after the first demyelinating event: Results from the multi-center SET study. PLoS One 2013;8(1):e53996.

76. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 2007;8:16.

77. Pittas F, Ponsonby AL, van der Mei IA, et al. Smoking is associated with progressive disease course and increased progression in clinical disability in a prospective cohort of people with multiple sclerosis. J Neurol 2009;256(4):577-85.

78. Sundstrom P, Nystrom L. Smoking worsens the prognosis in multiple sclerosis. Mult

Scler 2008;14(8):1031-5.

79. Gustavsen MW, Page CM, Moen SM, et al. Environmental exposures and the risk of multiple sclerosis investigated in a Norwegian case-control study. BMC Neurol

2014;14:196.

80. Gardener H, Munger KL, Chitnis T, Michels KB, Spiegelman D, Ascherio A.

Prenatal and perinatal factors and risk of multiple sclerosis. Epidemiology

2009;20(4):611-8.

81. Mikaeloff Y, Caridade G, Tardieu M, Suissa S. Parental smoking at home and the risk of childhood-onset multiple sclerosis in children. Brain 2007;130(Pt 10):2589-95.

101 82. Miller D, Barkhof F, Montalban X, Thompson A, Filippi M. Clinically isolated syndromes suggestive of multiple sclerosis, part I: Natural history, pathogenesis, diagnosis, and prognosis. Lancet Neurol 2005;4(5):281-8.

83. Chen JL, Wei L, Bereczki D, et al. Nicotine raises the influx of permeable solutes across the rat blood-brain barrier with little or no capillary recruitment. J Cereb Blood

Flow Metab 1995;15(4):687-98.

84. Minagar A, Alexander JS. Blood-brain barrier disruption in multiple sclerosis. Mult

Scler 2003;9(6):540-9.

85. Philbrick DJ, Hopkins JB, Hill DC, Alexander JC, Thomson RG. Effects of prolonged cyanide and thiocyanate feeding in rats. J Toxicol Environ Health

1979;5(4):579-92.

86. Smith KJ, Kapoor R, Hall SM, Davies M. Electrically active axons degenerate when exposed to nitric oxide. Ann Neurol 2001;49(4):470-6.

87. Coles AJ, Wing MG, Molyneux P, et al. Monoclonal antibody treatment exposes three mechanisms underlying the clinical course of multiple sclerosis. Ann Neurol

1999;46(3):296-304.

88. George J, Levy Y, Shoenfeld Y. Smoking and immunity: An additional player in the mosaic of autoimmunity. Scand J Immunol 1997;45(1):1-6.

89. Carlens C, Hergens MP, Grunewald J, et al. Smoking, use of moist snuff, and risk of chronic inflammatory diseases. Am J Respir Crit Care Med 2010;181(11):1217-22.

90. Costenbader KH, Kim DJ, Peerzada J, et al. Cigarette smoking and the risk of systemic lupus erythematosus: A meta-analysis. Arthritis Rheum 2004;50(3):849-57.

102 91. Prummel MF, Wiersinga WM. Smoking and risk of Graves' disease. JAMA

1993;269(4):479-82.

92. Majka DS, Holers VM. Cigarette smoking and the risk of systemic lupus erythematosus and rheumatoid arthritis. Ann Rheum Dis 2006;65(5):561-3.

93. Connor Gorber S, Schofield-Hurwitz S, Hardt J, Levasseur G, Tremblay M. The accuracy of self-reported smoking: A systematic review of the relationship between self- reported and cotinine-assessed smoking status. Nicotine Tob Res 2009;11(1):12-24.

103 CHAPTER 4: General Summary and Future Directions

The primary objective of this research was to examine the association between smoking and MS risk and MS progression through a systematic review and meta- analysis, and subsequently use the Bradford Hill criteria1 to further and more comprehensively evaluate causality. The results from this review suggest a statistically significant association between smoking and both MS risk and the progression from

RRMS to SPMS; however, the relationship between smoking and the conversion from

CIS to CDMS was statistically non-significant. Overall, there is strong evidence of a causal role of smoking in MS risk, and moderate evidence of a causal relationship between smoking and MS progression. However, it is unlikely that smoking plays an independent role in MS.

Rothman’s Model of Necessary and Sufficient Causes2 was used as the theoretical framework for this systematic review and meta-analysis as it offers a clear depiction of chronic disease causation. Past research suggests that MS is likely a multi-causal process in that it involves an interaction between a combination of genetic and environmental factors.3 As such, Rothman’s model is well suited to illustrate MS causation.

A depiction of two possible causal pathways involving MS risk factors is presented in Figure 4-1 (others may exist). Research has shown an interaction between smoking and the HLA-DRB1*15 allele, but only among those individuals without the more protective HLA-A*02 allele.4 In addition, a positive interaction between smoking and EBV was observed, but only among older individuals.5 Figure 4-1 illustrates two possible sufficient causes of MS (I, II), each consisting of different combinations of component causes (e.g., A, B, C, etc.) potentially including smoking. If a component

104

Figure 4-1. Potential causal pathways for MS using currently suspected component causes. Adapted from “Causes,” by K. J. Rothman, 1976, Am J Epidemiol, 104(6), p. 589. EBV = Epstein-Barr Virus; HLA = Human Leukocyte Antigen; ? = unknown component cause

105 cause is removed from one of the sufficient causes, such as having the protective HLA-

A*02 allele, being a younger age, or not smoking, the remaining components may become insufficient for causing MS. Given that not everyone who possess all the known component causes in either sufficient cause develops MS, it is very likely that other unknown yet to be discovered component causes also play a role in these sufficient causes (e.g., ? in Figure 4-1). Further research on MS risk factors is pertinent to the discovery of such component causes, as well as the discovery of other possible causal pathways in MS. For instance, in a pediatric study, a mere 5% of children without three

MS risk factors—HLA-DRB1*15, EBV, low vitamin D levels—had an MS diagnosis as opposed to 57% of children who had the three risk factors.6 While interactions between the risk factors were not found, the study itself highlighted the possibility of detecting the interactions with a larger sample. Overall, future research should include a more in depth investigation of the role of smoking in MS, while additional studies investigating interactions between risk factors are also needed. These investigations would be critical steps hopefully contributing to MS prevention.

In retrospect, this research provided a valuable opportunity to collate and extensively analyze a large number of studies relevant to the association between smoking and MS, enhancing the reliability of the results. Furthermore, a variety of research skills were acquired and practiced as a result of completing such an in depth meticulous and comprehensive review, including skills in database searching, compilation of evidence, critical thinking and analysis, and meta-analytic techniques. A limitation of this project was the involvement of only one primary reviewer, which may have resulted in a higher risk of errors in study selection and data extraction; however,

106 reliability checks and consultations to address uncertainties were done with a 2nd reviewer to limit these errors and increase both the internal and external validity of the review. In addition, there were times when the volume of work was overwhelming for one reviewer, and as such, certain steps in the review took longer than originally anticipated. Should the opportunity to conduct another systematic review and meta-analysis present itself, the ideal is to undertake such a project as part of a research team, both in terms of reducing bias and in terms of conducting the review in a more time-efficient manner.

107

4.1 References

1. Hill AB. The environment and disease: Association or causation? Proc R Soc Med

1965;58:295-300.

2. Rothman KJ. Causes. Am J Epidemiol 1976;104(6):587-92.

3. Jankosky C, Deussing E, Gibson RL, Haverkos HW. Viruses and vitamin D in the etiology of type 1 diabetes mellitus and multiple sclerosis. Virus Res 2012;163(2):424-30.

4. Hedstrom AK, Sundqvist E, Baarnhielm M, et al. Smoking and two human leukocyte antigen genes interact to increase the risk for multiple sclerosis. Brain 2011;134(Pt

3):653-64.

5. Salzer J, Stenlund H, Sundstrom P. The interaction between smoking and Epstein-Barr virus as multiple sclerosis risk factors may depend on age. Mult Scler 2014;20(6):747-50.

6. Banwell B, Bar-Or A, Arnold DL, et al. Clinical, environmental, and genetic determinants of multiple sclerosis in children with acute demyelination: A prospective national cohort study. Lancet Neurol 2011;10(5):436-45.

108

APPENDICES

109 APPENDIX A: Title, Abstract, and Full-Text Screening Questions Specific to Thesis

Title Yes No Unclear 1. Is the study an original Move to Question 2 Exclude Move to Question 2 research study? 2. Does the study investigate Move to Question 3 Exclude Move to Question 3 MS? 3. Does the study investigate Include and move to Exclude Include and move to smoking? Abstract Screen Abstract Screen Abstract Yes No Unclear 1. Is the study an original Move to Question 2 Exclude Move to Question 2 research study? 2. Does the study investigate Move to Question 3 Exclude Move to Question 3 MS? 3. Does the study investigate Include and move to Exclude Include and move to smoking? Full-Text Screen Full-Text Screen Full-Text Yes No Unclear 1. Is the study a peer- Move to Question 2 Exclude Discuss with reviewed original research supervisor study with a cross-sectional, cohort, or case-control design? 2. Does the study investigate Move to Question 3 Exclude Discuss with MS risk or progression as an supervisor outcome in participants (confirmed/self-reported diagnosis)? 3. Does the study investigate Move to Question 4 Exclude Discuss with smoking as an independent supervisor variable in participants? 4. Are MS patients being Include in review Move to Discuss with compared to a non-MS Question 5 supervisor healthy group or a general population group? 5. Are smokers being Include in review Exclude Discuss with compared to non-smokers? supervisor

110 APPENDIX B: Studies Included in Smoking-MS Risk Meta-Analysis and Corresponding Effect Estimates

First Author, Year Estimate(s) Provided Raw Data Estimate Calculated by Study by Reviewer* Al-Afasy, 2013 1.7 OR [0.9-3.4] Al-Shammri, 2015 1.9 OR [0.8-4.5] Alonso, 2011 1.72 OR [0.90-3.30] Antonovsky, 1965 Cases: 241 1.40 OR [1.05, 1.86] Exposed: 106

Controls: 964 Exposed: 347 Asadollahi, 2013 1.78 OR [1.22-2.59] Bjornevik, 2015 1.70 OR [1.40-2.08] Briggs, 2014 1.27 OR [1.03-1.58] Carlens, 2010 1.9 RR [1.4-2.6] Dahl, 2007 Cases: 250 1.10 OR [0.81, 1.49]* Exposed: 52

Controls: 371,878 Exposed: 71,772 Ghadirian, 2001 1.6 OR [1.0-2.4] Gustavsen, 2014 Cases: 518 2.41 OR [1.90, 3.06]* Exposed: 386

Controls: 903 Exposed: 495 Hedstrom, 2009 1.5 OR [1.3-1.8] Hedstrom, 2013 1.5 OR [1.4-1.7] Hernan, 2001 Past smoker: 1.2 RR [0.9- 1.6] Current smoker: 1.6 RR [1.2-2.1]† Hernan, 2005 1.3 OR [1.0-1.7] Jafari, 2009 1.09 OR [0.68-1.73] Kappus, 2015 1.91 OR [1.34-2.75] Kotzamani, 2012 Cases: 504 1.90 OR [1.49, 2.42] Exposed: 284

Controls: 591 Exposed: 239 Maghzi, 2011 2.67 OR [1.70-4.21] Mansouri, 2014 1.93 OR [1.31-2.73] Moccia, 2015 Cases: 265 1.05 OR [0.77, 1.42] Exposed: 99

Controls: 530 Exposed: 192

111 Mouhieddine, 2015 Cases: 248 0.97 OR [0.67, 1.40] Exposed: 99

Controls: 216 Exposed: 88 O'Gorman, 2014 1.9 OR [1.5-2.5] Ploughman, 2014 Cases: 743 0.54 OR [0.42, 0.69]* Exposed: 67

Controls: 8,795,596 Exposed: 1,372,885 Ramagopalan, 2013 1.32 OR [1.10-1.60] Riise, 2003 1.81 RR [1.13-2.92] Russo, 2008 Cases: 94 1.11 OR [0.51, 2.42] Exposed: 25

Controls: 53 Exposed: 13 Salzer, 2012 1.5 OR [0.98-2.3] Silva, 2009 2.0 OR [0.9-4.3] Simon, 2010 1.5 OR [1.0-2.4] Simon, 2015 1.4 OR [1.1-1.9] Simpson, 1966 Cases: 584 1.10 OR [0.93, 1.30]* Exposed: 333

Controls: 10,692 Exposed: 5848 Sundstrom, 2008 Cases: 91 1.32 OR [0.79, 2.20] Exposed: 55

Controls: 177 Exposed: 95 Thorogood, 1998 1-14 cigs/day: 1.2 RR [0.8-1.8] 15+ cigs/day: 1.4 RR [0.9-2.2]† Villard-Mackintosh, Ex-smoker: 1.5 RR [0.6- 1993 3.3] 1-14 cigs/day: 1.6 RR [0.8-3.1] 15+ cigs/day: 1.8 RR [0.8-3.6]† Zorzon, 2003 1.5 OR [0.9-2.4] OR = odds ratio; RR = risk ratio *Unadjusted, otherwise adjusted †Estimates initially combined in separate meta-analysis to derive average summary effect for input into main meta-analysis

112 APPENDIX C: GRADE Approach

Factor Description Grade Change 1. Study Limitations Limitations in observational Serious (-1)/very serious (-2) studies include: Selection bias, limitations in design and measurement bias (exposure execution (risk of bias) and outcome), lack of proper confounder control, loss to follow-up 2. Inconsistency Differences in effect estimates Serious (-1)/very serious (-2) across studies, which might be unexplained heterogeneity or due to different populations, inconsistency exposures, or outcomes 3. Indirectness Variability in populations, Serious (-1)/very serious (-2) exposures, comparators, and outcomes of interest to review authors and those available from the evidence 4. Imprecision Studies have a small sample Serious (-1)/very serious (-2) size and a small number of events, and therefore, CIs are wide; the 95% CI does not show an effect as well as conveys considerable benefit or harm 5. Publication Bias Researchers do not disclose Likely (-1)/high likelihood (-2) their studies, especially those that do not show an effect 6. Large Effect Estimate Large (RR > 2.0 or < 0.5) or Large (+1)/very large (+2) very large (RR > 5.0 or < 0.2) and consistent effect estimates are produced by observational studies of high methodological quality 7. Dose-Response Relationship Evidence of a gradient Present (+1) 8. Residual Confounding Possible confounders or biases Present (+1) would lower an observed effect, or produce a false effect when results do not show an effect CI = confidence interval; GRADE = Grading of Recommendations Assessment, Development, and Evaluation; RR = risk ratio Adapted from “What is “quality of evidence” and why is it important to clinicians?” by G. H. Guyatt et al., 2008, BMJ, 336(7651), p. 995-8; “GRADE guidelines: 3. Rating the quality of evidence,” by H. Balshem et al., 2011, J Clin Epidemiol, 64(4), p. 401-6; “GRADE guidelines: 4. Rating the quality of evidence – study limitations (risk of bias),” by G. H. Guyatt et al., 2011, J Clin Epidemiol, 64(4), p. 407-15; “GRADE guidelines: 6. Rating the quality of evidence – imprecision,” by G. H. Guyatt et al., 2011, J Clin Epidemiol, 64(12), p. 1283-93; “GRADE guidelines: 9. Rating up the quality of evidence,” by G. H. Guyatt et al., 2011, J Clin Epidemiol, 64(12), p. 1311-6.

113 APPENDIX D: Included Studies in Each Quantitative and Qualitative Analysis

1. Smoking and MS Risk (Meta-Analysis)

1. Al-Afasy HH, Al-Obaidan MA, Al-Ansari YA, et al. Risk factors for multiple

sclerosis in Kuwait: A population-based case-control study. Neuroepidemiology

2013;40:30-5.

2. Al-Shammri SN, Hanna MG, Chattopadhyay A, Akanji AO. Sociocultural and

demographic risk factors for the development of multiple sclerosis in Kuwait: A

case - control study. PLoS One 2015;10(7):e0132106.

3. Alonso A, Cook SD, Maghzi AH, Divani AA. A case-control study of risk

factors for multiple sclerosis in Iran. Mult Scler 2011;17(5):550-5.

4. Antonovsky A, Leibowitz U, Smith HA, et al. Epidemiologic study of multiple

sclerosis in Israel. I. An overall review of methods and findings. Arch Neurol

1965;13:183-93.

5. Asadollahi S, Fakhri M, Heidari K, Zandieh A, Vafaee R, Mansouri B. Cigarette

smoking and associated risk of multiple sclerosis in the Iranian population. J Clin

Neurosci 2013;20(12):1747-50.

6. Bjornevik K, Riise T, Cortese M, et al. Level of education and multiple sclerosis

risk after adjustment for known risk factors: The EnvIMS study. Mult Scler 2015.

doi:10.1177/1352458515579444.

7. Briggs FB, Acuna B, Shen L, et al. Smoking and risk of multiple sclerosis:

Evidence of modification by NAT1 variants. Epidemiology 2014;25(4):605-14.

**Duplicate data excluded

114 8. Carlens C, Hergens MP, Grunewald J, et al. Smoking, use of moist snuff, and risk

of chronic inflammatory diseases. Am J Respir Crit Care Med

2010;181(11):1217-22.

9. Dahl J, Myhr KM, Daltveit AK, Skjaerven R, Gilhus NE. Is smoking an extra

hazard in pregnant MS women? Findings from a population-based registry in

Norway. Eur J Neurol 2007;14(10):1113-7.

10. Ghadirian P, Dadgostar B, Azani R, Maisonneuve P. A case-control study of the

association between socio-demographic, lifestyle and medical history factors and

multiple sclerosis. Can J Public Health 2001;92(4):281-5.

11. Gustavsen MW, Page CM, Moen SM, et al. Environmental exposures and the

risk of multiple sclerosis investigated in a Norwegian case-control study. BMC

Neurol 2014;14:196.

12. Hedstrom AK, Baarnhielm M, Olsson T, Alfredsson L. Tobacco smoking, but not

Swedish snuff use, increases the risk of multiple sclerosis. Neurology

2009;73(9):696-701.

13. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Smoking and multiple sclerosis

susceptibility. Eur J Epidemiol 2013;28(11):867-74.

**Duplicate data excluded

14. Hernan MA, Jick SS, Logroscino G, Olek MJ, Ascherio A, Jick H. Cigarette

smoking and the progression of multiple sclerosis. Brain 2005;128(Pt 6):1461-5.

15. Hernan MA, Olek MJ, Ascherio A. Cigarette smoking and incidence of multiple

sclerosis. Am J Epidemiol 2001;154(1):69-74.

115 16. Jafari N, Hoppenbrouwers IA, Hop WC, Breteler MM, Hintzen RQ. Cigarette

smoking and risk of MS in multiplex families. Mult Scler 2009;15(11):1363-7.

17. Kappus N, Weinstock-Guttman B, Hagemeier J, et al. Cardiovascular risk factors

are associated with increased lesion burden and brain atrophy in multiple

sclerosis. J Neurol Neurosurg Psychiatry 2015.

doi:10.1136/jnnp-2014-310051.

18. Kotzamani D, Panou T, Mastorodemos V, et al. Rising incidence of multiple

sclerosis in females associated with urbanization. Neurology 2012;78(22):1728-

35.

19. Maghzi AH, Etemadifar M, Heshmat-Ghahdarijani K, et al. Cigarette smoking

and the risk of multiple sclerosis: A sibling case-control study in Isfahan, Iran.

Neuroepidemiology 2011;37(3-4):238-42.

20. Mansouri B, Asadollahi S, Heidari K, et al. Risk factors for increased multiple

sclerosis susceptibility in the Iranian population. J Clin Neurosci

2014;21(12):2207-11.

21. Moccia M, Lanzillo R, Palladino R, et al. The Framingham cardiovascular risk

score in multiple sclerosis. Eur J Neurol 2015;22(8):1176-83.

22. Mouhieddine TH, Darwish H, Fawaz L, Yamout B, Tamim H, Khoury SJ. Risk

factors for multiple sclerosis and associations with anti-EBV antibody titers. Clin

Immunol 2015;158(1):59-66.

23. O'Gorman C, Bukhari W, Todd A, Freeman S, Broadley SA. Smoking increases

the risk of multiple sclerosis in Queensland, Australia. J Clin Neurosci

2014;21(10):1730-3.

116 24. Ploughman M, Beaulieu S, Harris C, et al. The Canadian survey of health,

lifestyle and ageing with multiple sclerosis: Methodology and initial results. BMJ

Open 2014;4(7):e005718.

25. Ramagopalan SV, Lee JD, Yee IM, et al. Association of smoking with risk of

multiple sclerosis: A population-based study. J Neurol 2013;260(7):1778-81.

26. Riise T, Nortvedt MW, Ascherio A. Smoking is a risk factor for multiple

sclerosis. Neurology 2003;61(8):1122-4.

27. Russo C, Morabito F, Luise F, et al. Hyperhomocysteinemia is associated with

cognitive impairment in multiple sclerosis. J Neurol 2008;255(1):64-9.

28. Salzer J, Hallmans G, Nystrom M, Stenlund H, Wadell G, Sundstrom P. Smoking

as a risk factor for multiple sclerosis. Mult Scler 2013;19(8):1022-7.

29. Silva KR, Alvarenga RM, Fernandez YFO, Alvarenga H, Thuler LC. Potential

risk factors for multiple sclerosis in Rio de Janeiro: A case-control study. Arq

Neuropsiquiatr 2009;67(2A):229-34.

30. Simon KC, Schmidt H, Loud S, Ascherio A. Risk factors for multiple sclerosis,

neuromyelitis optica and transverse myelitis. Mult Scler 2015;21(6):703-9.

31. Simon KC, van der Mei IA, Munger KL, et al. Combined effects of smoking,

anti-EBNA antibodies, and HLA-DRB1*1501 on multiple sclerosis risk.

Neurology 2010;74(17):1365-71.

**Duplicate data excluded

32. Simpson CA, Newell DJ, Schapira K. Smoking and multiple sclerosis. Neurology

1966;16(10):1041-3 passim.

117 33. Sundstrom P, Nystrom L, Hallmans G. Smoke exposure increases the risk for

multiple sclerosis. Eur J Neurol 2008;15(6):579-83.

34. Thorogood M, Hannaford PC. The influence of oral contraceptives on the risk of

multiple sclerosis. Br J Obstet Gynaecol 1998;105(12):1296-9.

35. Villard-Mackintosh L, Vessey MP. Oral contraceptives and reproductive factors

in multiple sclerosis incidence. Contraception 1993;47(2):161-8.

36. Zorzon M, Zivadinov R, Nasuelli D, et al. Risk factors of multiple sclerosis: A

case-control study. Neurol Sci 2003;24:242-7.

2. Risk Factor Interactions (Qualitative)

1. Briggs FB, Acuna B, Shen L, et al. Smoking and risk of multiple sclerosis:

Evidence of modification by NAT1 variants. Epidemiology 2014;25(4):605-14.

2. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Alcohol as a modifiable

lifestyle factor affecting multiple sclerosis risk. JAMA Neurol 2014;71(3):300-5.

**Duplicate data excluded

3. Hedstrom AK, Sundqvist E, Baarnhielm M, et al. Smoking and two human

leukocyte antigen genes interact to increase the risk for multiple sclerosis. Brain

2011;134(Pt 3):653-64.

**Duplicate data excluded

4. Salzer J, Stenlund H, Sundstrom P. The interaction between smoking and

Epstein-Barr virus as multiple sclerosis risk factors may depend on age. Mult

Scler 2014;20(6):747-50.

**Duplicate data excluded

118 5. Simon KC, Schmidt H, Loud S, Ascherio A. Risk factors for multiple sclerosis,

neuromyelitis optica and transverse myelitis. Mult Scler 2015;21(6):703-9.

6. Simon KC, van der Mei IA, Munger KL, et al. Combined effects of smoking,

anti-EBNA antibodies, and HLA-DRB1*1501 on multiple sclerosis risk.

Neurology 2010;74(17):1365-71.

7. Sundqvist E, Sundstrom P, Linden M, et al. Lack of replication of interaction

between EBNA1 IgG and smoking in risk for multiple sclerosis. Neurology

2012;79(13):1363-8.

**Duplicate data excluded

3. Smoking and Progression from CIS to CDMS (Meta-Analysis)

1. Arikanoglu A, Shugaiv E, Tuzun E, Eraksoy M. Impact of cigarette smoking on

conversion from clinically isolated syndrome to clinically definite multiple

sclerosis. Int J Neurosci 2013;123(7):476-9.

2. Di Pauli F, Reindl M, Ehling R, et al. Smoking is a risk factor for early

conversion to clinically definite multiple sclerosis. Mult Scler 2008;14(8):1026-

30.

3. Kuhle J, Disanto G, Dobson R, et al. Conversion from clinically isolated

syndrome to multiple sclerosis: A large multicentre study. Mult Scler

2015;21(8):1013-24.

4. Smoking and Progression from CIS to CDMS (Qualitative)

1. Correale J, Farez MF. Smoking worsens multiple sclerosis prognosis: Two

different pathways are involved. J Neuroimmunol 2015;281:23-34.

119 2. Horakova D, Zivadinov R, Weinstock-Guttman B, et al. Environmental factors

associated with disease progression after the first demyelinating event: Results

from the multi-center SET study. PLoS One 2013;8(1):e53996.

5. Smoking and Progression from RRMS to SPMS (Meta-Analysis)

1. Healy BC, Ali EN, Guttmann CR, et al. Smoking and disease progression in

multiple sclerosis. Arch Neurol 2009;66(7):858-64.

2. Hernan MA, Jick SS, Logroscino G, Olek MJ, Ascherio A, Jick H. Cigarette

smoking and the progression of multiple sclerosis. Brain 2005;128(Pt 6):1461-5.

3. Koch M, van Harten A, Uyttenboogaart M, De Keyser J. Cigarette smoking and

progression in multiple sclerosis. Neurology 2007;69(15):1515-20.

4. Roudbari SA, Ansar MM, Yousefzad A. Smoking as a risk factor for

development of secondary progressive multiple sclerosis: A study in Iran, Guilan.

J Neurol Sci 2013;330(1-2):52-5.

5. Sundstrom P, Nystrom L. Smoking worsens the prognosis in multiple sclerosis.

Mult Scler 2008;14(8):1031-5.

6. Smoking and Progression from RRMS to SPMS (Qualitative)

1. Correale J, Farez MF. Smoking worsens multiple sclerosis prognosis: Two

different pathways are involved. J Neuroimmunol 2015;281:23-34.

2. Pittas F, Ponsonby AL, van der Mei IA, et al. Smoking is associated with

progressive disease course and increased progression in clinical disability in a

prospective cohort of people with multiple sclerosis. J Neurol 2009;256(4):577-

85.

120 7. Passive Smoke Exposure and MS Risk (Qualitative)

1. Gardener H, Munger KL, Chitnis T, Michels KB, Spiegelman D, Ascherio A.

Prenatal and perinatal factors and risk of multiple sclerosis. Epidemiology

2009;20(4):611-8.

2. Gustavsen MW, Page CM, Moen SM, et al. Environmental exposures and the

risk of multiple sclerosis investigated in a Norwegian case-control study. BMC

Neurol 2014;14:196.

3. Hedstrom AK, Baarnhielm M, Olsson T, Alfredsson L. Exposure to

environmental tobacco smoke is associated with increased risk for multiple

sclerosis. Mult Scler 2011;17(7):788-93.

4. Hedstrom AK, Bomfim IL, Barcellos LF, et al. Interaction between passive

smoking and two HLA genes with regard to multiple sclerosis risk. Int J

Epidemiol 2014;43(6):1791-8.

**Duplicate data excluded

5. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Smoking and multiple sclerosis

susceptibility. Eur J Epidemiol 2013;28(11):867-74.

**Duplicate data excluded

6. Mikaeloff Y, Caridade G, Tardieu M, Suissa S. Parental smoking at home and the

risk of childhood-onset multiple sclerosis in children. Brain 2007;130(Pt

10):2589-95.

7. Ramagopalan SV, Lee JD, Yee IM, et al. Association of smoking with risk of

multiple sclerosis: A population-based study. J Neurol 2013;260(7):1778-81.

121 8. Sundstrom P, Nystrom L, Hallmans G. Smoke exposure increases the risk for

multiple sclerosis. Eur J Neurol 2008;15(6):579-83.

8. Maternal Prenatal Smoke Exposure and MS Risk (Qualitative)

1. Gardener H, Munger KL, Chitnis T, Michels KB, Spiegelman D, Ascherio A.

Prenatal and perinatal factors and risk of multiple sclerosis. Epidemiology

2009;20(4):611-8.

2. Montgomery SM, Bahmanyar S, Hillert J, Ekbom A, Olsson T. Maternal

smoking during pregnancy and multiple sclerosis amongst offspring. Eur J

Neurol 2008;15(12):1395-9.

3. Mueller BA, Nelson JL, Newcomb PA. Intrauterine environment and multiple

sclerosis: A population- based case-control study. Mult Scler 2013;19(1):106-11.

4. Ramagopalan SV, Lee JD, Yee IM, et al. Association of smoking with risk of

multiple sclerosis: A population-based study. J Neurol 2013;260(7):1778-81.

122 FirstAuthor,Author, FF/M/MSexRatio Ascertainmentof MSDiagnosticDiagnostic YearYear StudyStudyD Designesign ObjectiveObjective CountryCountry SampleSampleSi Sizeze Source/DiseaseSource/DiseaseHi Historystory(C (Casesasesor C Cohort)ohort) Source/DiseaseSource/DiseaseHi Historystory(C (Controls)ontrols) (case/control)(case/control) AgeAge (case/control)(case/control) SmokingSmokingSt Statusatus SmokingSmoking MSC Courseourse(%) MethodMethod ConfoundersConfoundersM Matchedatchedor AdjustedAdjusted MSRisk n=101n=101cases/n cases/n=202=202 MS Association of KuwaitKuwait (all newlynewly diagnosed UniversityUniversity studentsstudents and governmentgovernment employees/noemployees/no hxhx 33.86 (9 .1 SD)/33 .87 EverEversmoked QuestionnaireQuestionnaire in AlͲ AfasyAfasy,,2013 CaseͲcontrol To ascertain risk factors of MS Kuwait 1.2/1.21.2/1.2 MS (100) Neurologist Age,Age,gender gender,,nationality controls patientspg)/pregistered)/confirmed)/ f MSpatients offMSorneurologicalg disorder ()(9.2SD) (Conservative)() interview n=195cases/ Age,sex , BMI, ethnicity, in Kuwait at TodetermineassociationbetweenppotentialotentialsocioͲ 27Ͳ 38,38,33.2(10.4(10.4 AlͲ ShShammri, i n=142142controls t lwith ith MMubarak b k AlͲKabir K biSpecialist S i li t HHospital's it l' MS KuwaitK itcommunity itrandom dselection/healthy l ti /h lthno MS EkdEversmoked QuestionnaireQtiiiin RRMS(85); (85)SPMS (13); (13) titimeof f GGulf lf WWar, MS ffamily il hhx, CaseͲcontrol demographic and environmental MS risk factors Kuwait 16/181.6/1.8 SD)/<30to >40, 33.8 McDonald (2010) APPENDIX E:DetailsofStudiesIncludedinMeta-Analysis 2015 smoking data (out of clinic/MS patients diagnosis (NonͲ Conservative) interview PPMS(2) related disease hx,hx,sun exposure exposure,, anddMS ()(9.5SD) 146contro controls)ls) preferredpreferred outdooroutdoor clothingclothing RRMS (84); SPMS with Age,g,ggender, ,sunexposure, p ,MSfamily y ToTo determinedetermine associationassociation betweenbetween lifestylelifestyle andand nn= 394cases cases/n/n=394 MS clinicsclinics inin TehranTehran, IsfahanIsfahan, MashhadMashhad, BandarBandar 30.9 (8 .9 SD)/31 .0 EverEversmo smokedked QuestionnaireQuestionnaire inin rerelapselapse(14) (14);;SPMS Alonso, 2011 CaseͲcontrol Iran Childhood friend or relative of case/no MS 37/163.7/1.6 McDonald (2001) hx, past surgeries, infectious disease environmentalfactorsandMSrisk controls AbbasAbbas,,ShirazShiraz/consecutive/consecutiveMSppatientsatients ((10.110.1SDSD)) ((Conservative)Conservative) interview withoutrelapsesrelapses((0.3);0.3); hhx PPMS (2);(2);PRMS (0 .3) AntonovskAntonovskyy,, TocomparecompareMSppatientsatientswithcontrolswithrespectrespect n=241n 241casescases/n/n=964 964 Eversmoked AuthorͲdevelo developedped ClCaseͲcontrol IlIsrael NiNationwide idsurvey/MSyp /MSipatients PopulationPlipgyregistry iof fIlidIsraelviarandomselection li SexShdͲmatchedAhd Ageg Ͳmatched QiQuestionnaire i MS (100)() AifiiAge,g,sex, ,region goforigin g 1965 toto exposuret toopo potentialtential MS causalcausal factorsfactors concontrolstrols (Conservative)(Conservative) diagnosticdiagnostic criteriacriteria IranianMSRegistrygycomprised pofpatients p from <30toш50,,(31(10.0 Medicalrecords Age,g,sex, ,education, ,ethnicity, y, To determinedeterminere relationshiplationship betweenbetween environmentalenvironmental nn= 662cases cases/n/n=39 3944 FriendFriend or neighbourneighbour ofof case/nocase/no MSMS or diagnosisdiagnosis ofof Eversmo smokedked Poser( (1983)1983), Asadollahi, 2013 CaseͲcontrol Iran public neurology departments in 27 Iranian 38/163.8/1.6 SD)/<30 toш50,33 .6 (case),interview MS (100) birthplace, area of residence, sun tobaccosmokeandriskofMS controls otherneuroloneurologicalgicaldisorders ((Conservative)Conservative) McDonald((2001)2001) cities/definitei i /d fi i MSppatients i (8(8.9() 9SD) (l)(control)() exposure,p,ifinfectious i didisease hhx,, Age,Age,sex sex,,education education,, infectious EnvironmentalFactorsinMS(EnvIMS)()gNorwegian To determinedetermine relationshiprelationship betweenbetween educationeducation andand nn= 953953cases cases/n/n=1717 1717 EnvIMSEnvIMS NorwegianNorwegian studystudyͲͲ PopulationPopulation registryregistry viavia Eversmo smokedked mononucleosismononucleosis,outdooroutdoor timetime, Bjornevik,2015 CaseͲcontrol Norway studyͲͲMS registry, Bergen biobank (enrolls SexͲmatched AgeͲmatched Questionnaire MS (100) McDonald (2005) MSafterMSriskfactoradjustmentadjustment controls randomselection/healthyselection/healthy ((Conservative)Conservative) vitaminDsusupplementationpplementation,fattfattyyfish subjectsbjjy ffromentire icountry)/MS )/MS)ppatients i consumptionconsumption, bodybody sizesize KaiserPermanenteMedicalCarePlan()(KPNC) 18Ͳ69 ,50 50sson average members identified through medical records (25Ͳ Birthyear ,sex ,education ,ancestry TodetermineifggeneenevariationamonamonggnonͲ HLA N.CaliforniaN.California,, n=1012n 1012cases/ncases/n=576 576 RandomlRandomlyyselected KPNCmembersfrommedical ((bornborn19581958,,8.7 Eversmoked BBriggs, igg 2014 Cl CaseͲcontrol 30%of fanorthern h CCalifornia lif iservice iregiong i 4/54/5.3 3 IiInterview MS (100)() McDonaldM D ld (2005)() (similar(i(f ilresults lobtained b i dffrom genesa alterslters effecteffect ofof smokingsmokingon riskrisk ofof MS USA controlscontrols records/norecords/no diagnosisdiagnosis ofof MS or similarsimilar diseasesdiseases SD)/50s50s on average (Non(NonͲ Conservative)Conservative) population included,included, representing over 3 million matched analysis) (born(,)1957,8.1SD) people)/MSpeople)/MS patientspatients <29to 50+, 42.4 (male MSAssociationofMontrealEast,neuroloneurologistgistand TdTodetermine iassociation ii bbetweensocio iͲ mean),)377(f)(37.7(female l MontrealMontreal, n=200n=200cases cases/n=202/n=202 GPre referralsferrals,genera generall publicpublic( (newspapernewspaper,ra radiodio EverEversmo smokedked QuestionnaireQuestionnaire inin AgeAge,sex, sexarea o offres residenceidence, Ghadirian,Ghadirian, 2001 CaseͲcontrol demographics,demographics, lifestyle factors,factors, and medical history General population via random digit dialling 2.2/2.22.2/2.2 mean)/<29to 50+,50+, MS (100) Notstated Quebec,Q,Canada controls announcements)/newly)/ ydiagnosed g incidentMS (Conservative)() interview education 12 andand MSMS riskrisk 41.2 ( (malemalemean mean)),38 .1 patients ((femalefemalemeanmean))

3 n=518518caseswith i h smoking data (out of <39to>52,,(50.5(12.7 To determinedetermine relationshiprelationship betweenbetween lifestylelifestyle andand 530 cases)/cases)/ OsloOslo MSMS RegistryRegistry (patients(patients fromfrom largestlargest NorwegianNorwegian BloodBlood donorsdonors fromfrom NationalNational Bone Marrow Donor Eversmo smokedked Poser( (1983)1983), Gustavsen,2014 CaseͲcontrol Norway 28/142.8/1.4 SD)/<39to >52, 44(6 .9 Questionnaire RRMS (89);other (11) None environmentalfactorsandMSrisk n=903n 903controls clinicrecruited)/MSrecruited)/MSppatientsatients ReRegistrygistryviarandomselectionselection/healthy/healthy (NonͲ Conservative)Conservative) McDonald((2010)2010) SD)) withwith smokingsmoking datadata (out()of918controls) To determine in what way age at initial smoke Age,gender ,area of residence , n=6085n 6085 GenesandEnvironmentinMS((GEMS)GEMS)ͲͲPatients exposure,pgysmokingkiintensity, i i dduration, i ddose,and d EkdEversmoked ancestry,yppassiveiksmokeexposurep HedstromHedstrom,2013 CaseCaseͲcon controltrol SwedenSweden casescases/n=5357/n=5357 fromfrom SwedishSwedish NationalNational MS Registry/prevalentRegistry/prevalent MS GEMSͲͲ PopulationPopulation registerregister viavia randomrandom selectionselection 26/292.6/2.9 16Ͳ70/16 Ͳ70 QuestionnaireQuestionnaire MS (100) McDonaldMcDonald cessation impact relationship between smoking (Conservative) (similar results obtained from controls patientsp andand riskrisk ofof MSMS matchedmatched analysis)analysis) Epidemiological Investigation of MS (EIMS)ͲͲ Age,sex ,area of residence ,ancestry Todeterminerelationshiprelationshipbetweensmokinsmokinggand n=902n 902casescases/n/n=1855 1855 Eversmoked HHedstrom, d2009 CaseClͲcontrol SdSweden PiPatientsfrom fneurology lgyclinics lii((participation (p ii p i ffrom EIMSͲͲ PPopulationpg l iregister ivia irandom dselection l i 2.6/2.626/26 16Ͳ70/16 Ͳ70 QiQuestionnaire i MS (100)() McDonaldMD ld (similar(i(f ilresults lobtained b i dffrom SwedishSwedish snuffsnuff andand MS riskrisk controlscontrols (Conservative)(Conservative) Swedish university hospitals)/incident MS patients matched analysis) GeneralPractice Research Database (GPRD) (over <<3030to ш50,50,36 36.0.0(9 (9.1.1 Age,Age,sex sex,,GP practice practice,,date of NestedcaseͲ Todeterminerelationshippgbetweensmokingand n=201cases/n=1913/ Eversmoked RRMS(79);()();SPMS(11); ()(); Poser(1983)()blindto HernanHernan,2005 BritainBritain 3m millionillion BritishBritish citizenscitizens included)/confirmedincluded)/confirmed MS GPRD viavia randomrandom selection/noselection/no MS diagnosisdiagnosis 24/242.4/2.4 SD)/<30SD)/<30 toшш50,36 .0 MedicalMedical recordsrecords enrollingenrolling inin practicepractice, smokingsmoking control MS risk controls (Conservative) PPMS(10) exposuredata ppatientsatients ((9.19.1SDSD)) informationavailabilitavailabilityy Todetermineassociationbetweensmokinsmokinggand n=136n 136casescases/n/n=204 204 MembersofmultiplexmultiplexfamiliesfromErasMSclinic SiblinSiblingsgsofcasesfrommultimultiplexplexfamilies/healthyfamilies/healthywith 30Ͳ 81,81,52.6((11.411.4SDSD)/)/ Eversmoked RelaRelapsingpsingonset((78);78); JJafari, f i 2009 Ctl CaseͲcontrol NthlNetherlands d 17/081.7/0.8 QtiiQuestionnaire MMcDonald D ld (2001)() AAge,g sex MS riskrisk controlscontrolsoorro otherther clinics/MSclinics/MS patientspatientsnnoo reportedreported signssigns ofof MS 22Ͳ79 ,52 .8 (11 .6 SD) (Conservative)(Conservative) progressiveprogressiveonse onsett(22)

n=504504cases <25to ш65,44 .1 (13 .0 withsmokingsmokingdata MS EEpidemiologypgygj id i l PProgram PProject jof f CCreteused dto SD))((males), ( l )), 4242.9 9 (12(12.2( 2 ToTo determinedeterminepo potentialtential environmentalenvironmental factorsfactors (out(out ofof 657 cases)/cases)/ EverEversmo smokedked McDonaldMcDonald (2001, Kotzamani,Kotzamani,2012 CaseͲcontrol Crete,Crete,Greece locateevery new MS subject on island/new MS Local population via random selection/healthy 1.6/1.61.6/1.6 SD) (females)/44.1(females)/44.1 Questionnaire MS (100) Age,Age,gender gender,,area of residence relatedtoincreasedMSincidence n=591n 591controls ((Conservative)Conservative) 20052005)) cases (13(13.1 1SD SD)) (ma les ), 4343.6 6 with smoking data ((1212.6.6SDSD))((females)females) ((out()of f593 controls) l ) TdtTodetermine ire ltihibtlationshipbetween diffdifferent t n=516516cases/ n=1090 1090 IfhIsfahan MSSMSSociety it (l(close ttoa llpeop lewit ithMSihMSin 3232.75 75( (88.68 68SD SD)/32)/32.86 86 EkdEversmoked RRMS (83)(83); SPMS (14)(14); MMcDona Dld (2001(2001, Maghzi, 2011 CaseͲcontrol Isfahan, Iran Siblings of cases/healthy no MS 37/093.7/0.9 Questionnaire Age,sex smokinsmokinggbehavioursandMSrisk controls Isfahanreregistered)/MSgistered)/MSppatientsatients ((99.65.65SD)SD) ((Conservative)Conservative) PPMS((3)3) 20052005)) n=1217n=1217cases Age,Age,sex sex,,ethnicity ethnicity,,place of birth birth,, witht ssmoking o g data QuestQuestionnaire o a ein areao off residenceresidence,e edducat cationion,medicalmedical To determine relationship between potential MS (out of 1403 cases)/ University hospital MS clinics located across 27 15Ͳ59,32 .6 (8 .7 Current smoking interviewͲͲ RRMS (78); SPMS (17); Poser(1983) , Mansouri,2014Case Ͳcontrol Iran HosHospitalpitalstaffstaff/no/nohxofMSorneurologicalneurologicaldisorder 3.13.1/3/3.3.3 hx,autoimmunedisorderfamilfamilyyhx, riskikffactorsand dMSMS n=787787controls l IranianI icities/CDMS i i /CDMSpatientsp i SD)/32SD)/32.3)( 3 (9(9.7 7SD) ) (Conservative)(C() i ) ii interviewers blidblindto PPMS(3); (3)();PRMS (2) () McDonaldM D ld (2005)() vaccinationsvaccinations,pastpastsurger surgeriesies,sun with smoking data purpose,purpose,design exposureeposue (out tof 883contro tl s) To determine CV riskdifferencesbetweenMS 18Ͳ65, 65,42 42.2.2(10 (10.9.9 n=265cases/n=530/ FedericoIIUniversityypHospital'sMS FedericoIIUniversityypHospitalvisitorstooccupationalp Currentsmokingg Poser(1983),()(), MocciaMoccia, 2015 CaseCaseͲcontro controll patientspatients andand controlscontrols andand associationassociation betweenbetween CV NaplesNaples, ItalyItaly 16/161.6/1.6 SD)/20Ͳ 65, 42. 9 (9. 7 InterviewInterview RRMS (81);(81);SPMS (19) AgeAge,gen genderder controls centre/consecutive definite MS patients medicine branch/no MS (NonͲConservative) McDonald (2010) riskandMSoutcomesinppatientsatients SDSD)) n=248n=248cases ToascertainEBVp,prevalence,aswellas pprevalence withsmokinggdata 14Ͳ69 ,37 .8 (12 .1 Mouhieddine, of other MS risk factors in MS patients and (out of 249 cases)/ AmericanUniversity of Beirut Medical Center 'sMS sMS Participants from general population in vitamin D study Current smoking Notstated (self Ͳ CaseͲcontrol BeirutBeirut,,Lebanon 1.51.5/1.7/1.7 SDSD)/14)/14Ͳ69 69,,46.4((12.512.5 RRMS((79);79);other((21)21) Notstated Gender 2015 controls,t land d ddetermine t i ifanti tiͲ EBVantibody tib d tittitres n=216216controls t l center/MSt /MSpatients ti t attAAmerican iUniversity U i itof fBBeirut i tMMedical di lCCenter/no t / MS (N(NonͲ CConservative) ti ) report)t) SD) areassociated with such factors with smoking data (out(out ofof 230 controls)controls) McDonald (2005) (G(Group 1)1);casenotes t To determine relationship between smoking and Queensland, n=560n=560cases/n cases/n=480=480 Gold Coast HospitalHospital's's MS clinic (Group 1) and MS Mandatory electoral roll of population older than 18 Eversmoked (Non Ͳ RRMS (53); SPMS (30); OOGorman'Gorman,,2014 CaseͲcontrol 4.5/1.94.5/1.9 50 (12SD)/57 (15SD) Questionnaire recordedby GP Age,Age,sex sex,,ethnicity ethnicity,,area of residence MSriskk Australial controlsl Societyy(()/(Group p2)/MS )/ ppatients y/years/no/dMSdiagnosis greported p d Conservative)) PPMS()();()(14);CIS(4) () and/orand/or neurologistneurologist ((GroupGroup22)) CdiCllbtiPjtGtiCanadianCollaborativeProjectonGenetic To determine association between individual RamaRamagopalangopalan, n=3157cases/n=756cases/n=756 SusceSusceptibilityptibilitytoMS((CCPGSMS)CCPGSMS)comcomprisedprisedof 48.048.0((1313.9.9SDSD)/45)/45.4.4 Eversmoked((NonNonͲ ClCaseͲcontrol smokingkigghabits h biand dmaternal lsmoking ki dduring i g CdCanada SfSpousep ofcase 3.3/0.333/03/ QiQuestionnaire i RRMSS(9)(91);();other h (9)() Poser(()(1983) 983) Age,g,sex 2013 controlscontrols patientspatients fromfrom 16 MS clinicsclinics across 8 CanadianCanadian (13.9 SD) Conservative)Conservative) pregnancy and MS risk pprovinces/MSrovinces/MSppatientsatients To determineif levels of homocysteine are hihheightenedggpp damongMS Spatients i compareddto ReggioReggioCalabria, n94n=94cases/n=53cases/n 53 BlooddonorsfromThrombosisandHemostasis 20Ͳ 69,36.63(10.36 Currentsmokinsmokingg RRMS(82);SPMS(16); RRusso, 2008 Ctl CaseͲcontrol NNeurol ogyc lilinic ire ferra ls /consecutive ti MSpatients ti t 12/111.2/1.1 ItInterview i MMcDona Dld (2005)(2005) AAge controls and whether hyperhomocysteinemia is Italy controls Centre/healthy SD)/37.15 (12.06 SD) (NonͲConservative) PPMS(2) reltdtlatedtocognitive iti iimpairment i tamong MS MS databasedatabasef fromromreg registriesistriesw withith ICD codescodes crossͲ Cotinine: Current Cotinine: Cotinine: nn=192=192 linkedtoN.SwedenHealth&DiseaseStudyy smokingki gш10 Cotinine: NestedNestedcase Ͳ cases/ncases/n=384 Cotinine:Cotinine: 16Ͳ 60, 26 Cotinine:Cotinine: RRMS (63);(63); Cotinine:Cotinine: AgeAge bloodblood sampledsampled, sexsex, To determine relationship between smoking and Northern (NSHDS) Cohort (incl.blood samples gatheredin Cotinine: ng/ml Immunoassay; SelfͲ SalzerSalzer,,2012 controlcontrol;; controlscontrols;; NSHDSandNSMCcohorts ((median)/16median)/16Ͳ60 60,,26 SPMS((23);23);PPMS/PRMSPPMS/PRMS McDonald((2010)2010) biobankbiobank,,bloodsamsamplingplingdatedate;;SelfͲ MSrisk i k SdSweden hhealth lthprogramsstart t t 1985)and dNSN.Sweden d 1111.8/11.8 8/11 8 (Conservative)(Conservative); report:t SelfSelfͲ report:report: SelfSelfͲrepor report:t:n= n=141141 (median)(median) (12) report:report:MS onse onsettage, agesex Maternity Cohort (NSMC) (incl.(incl. blood samples SelfͲ report:report:Ever Questionnaire CaseͲcontrol cases/n=254/ controls gatheredgathered inin pregnancy start 1975)/MS patientspatients smokedsmoked

124 Rio de Janeiro,Janeiro, nn=81=81cases/n=81 cases/n=81 Hospital da Lagoa' s Neurology 39.539.5 (11.3(11.3 SD)/38.5SD)/38.5 Current smoking Questionnaire in Silva,l ,2009 CaseͲcontroll TostudydkffyriskfactorsforMS NonͲrelativelfdfriendorneighbourg hbfofcase 2.1/2.1/ MS(100)() Poser(1983)() Age,g,ggender,dl ,pplaceof fbhbirth BrazilBrazil controlscontrols Department/consecutiveDepartment/consecutiveCDMS patientspatients (11.3 SD) (Non(NonͲ Conservative)Conservative) interviewinterview AcceleratedCureProjectProjectreferralsofcase((relatedrelatedand TTo ddetermine t irelationship l ti hi bbetween tpotential t ti l MS WlthWaltham, CtCasereport n=1190n=1190cases/n cases/n=454=454 AcceleratedCure Project subjects recruited from unrelated)/healthy with no evidence of CNS Eversmoked Adapted from Age,gender ,race/ethnicity , study Simon,Simon, 2015 CaseͲcontrol riskfactors and MS,MS, neuromyelitis optica,optica, and Massachusetts,Massachusetts, 3.5/1.83.5/1.8 46 (11SD)/48 (17SD) document in MS (100) controlsl MSclinicslbornearbyy/pcommunities/MS /patients dldemyelinationygggygsuggestingMSandddlnodemyelinating ()(Conservative) McDonaldld site transversetransversemye myelitislitis USA interviewinterview disorder diagnosis n=135135cases To determine interaction between smoking, antiͲ TTasmaniaasmania,, withsmokingdata MSMSsocsocietiesieties,,referralsreferralsfrfromom 4433.9.9((mean)/43mean)/43.9.9 EverEverssmokedmoked RRMRRMSS(65);SSPMSPMS((27);27); MRIccriteriariteria,,PPoseroser Simon,,2010 CaseͲcontroll EBNAantibodybdytitresand dHLAͲDR15asitrelatesl to Communityyymembersbffrommandatory delectoral l lllroll 2.1/2.1/ Interview Age,g,sex AustraliaAustralia (out(out ofof 136 cases)/cases)/ neurologists/prevalentneurologists/prevalent MS patientspatients (mean)(mean) (Conservative)(Conservative) PPMS(8) (1983) riskof MS n=272n 272controls n=584584 EidEvidenceo flflesion i To determine relationship between smoking and Datafrom MS surveyin northern Data from British Tobacco Manufacturers Standing Current smoking SimSimpsonpson,1966 Case Ͳcontrol N.N.EnglandEngland casescases/n/n=10 10,692 1.51.5/1/1.1.1 16Ͳ60+ 60+/16/16Ͳ60+ Interview MS((100)100) disseminationin None MSrisk i k EEngland/probableg/p(p l d/ b bl ((notpossible) ibl ))p MSpatients i CommitteeC inational i lsurveyy( (1958) ) (N(Non()Ͳ CConservative) i ) controlscontrols spacean andd timetime Cotinine: Current Nationalat o aMS S databasefrom o susurvey eylinked ed to Cotinine:Cot e: CCotinine: ti in=109 109 smokkiingшш 10 NSHDS Cohort (includes blood samples from Immunoassay on Vasterbotten casescases/n/n=218 218 Cotinine:11.7.7/1/1.7.7 Cotinine:30Ͳ 68,50 ng/mlng/ml((NonNonͲ TdTodetermine irelationship li hibpgbetweensmoking kiand d VVasterbotten b IIntervention i PProgramg start19801980s bliblinded d d blblood d SundstromSundstrom, 2008 CaseCaseͲcon controltrol CountyCounty, controls;controls; NSHDS CohortCohort SelfSelfͲ report:report: (median)/30(median)/30Ͳ 68, 50 Conservative);Conservative); MS (100) PoserPoser(1983) AgeAge,sex, sexbl bloodood collectioncollection year MS risk invitingall Vasterbotten residents aged 40,40, 50,50, 60 samples; SelfͲ N.Sweden.S ede SelfSeͲreport: epo t:n=91 9 1.3/1.2.3/ . (median)(eda) SelfͲ report:report:Ever y)l),also MONICAChtMONICACohortan dMdMammograp hy report:t cases/n=177 controls smoked ScreeninScreeninggCohort((MSC)/MSMSC)/MSppatientsatients QQuestionnaireuestionnaire (C(Conservative)() i ) Questionnaire in 17Ͳ65,,(42.1(10.9 To determinedetermine associationassociation betbetween een potentialpotential MS nn=140 140 cases/ncases/n=131 131 UnUniivers ersitity ofof Trieste'sTrieste's MS centre/conseccentre/consecut tiive e MS 18/Se1.8/SexͲ EEver ersmo smokedked interinterv ieiewͲͲ RRMS (71)(71); SPMS (19)(19); Zorzon, 2003 CaseͲcontrol Trieste, Italy Blood donors from local blood transfusion centre SD)/AgeͲmatched(±2 McDonald (2001) Age,sex riskfactorsandMSrisk controls ppatientsatients matched ((Conservative)Conservative) interviewersblindto PPMS((9)9) years)y)) hypothesishypothesis, designdesign n=240,318, everͲ smokers and nonͲ tobaccob(fusers((of ToTo determinedetermine thethe extentextent toto whichwhich smokingsmoking andand 277,777)/277,777)/ Record linkagewith Prospectivep moistsnuffareassociatedwithMS,,rheumatoid Constructionworkers/MS/phospitalizationsprior pto Eversmoked Age,g,areaofresidence,,moistsnuff CarlensCarlens,2010 SwedenSweden 118787d developedeveloped MSMS N/AN/A MaleMale<<2424to шш5555, 36( (mean)mean) MedicalMedical recordsrecordsMMSS (100) SwedishSwedish HospitalHospital cohort arthritis, ulcerative colitis, Crohn' s disease, and baseline visit excluded from analysis (Conservative) use whowereeverͲ DischarDischargegeReRegistergister sarcoidosisid irisk i k smokerssmokersor non Ͳ tobaccousers((of214 MS patients) NHSNHS:30 Ͳ55; 55; 42.842.8(mean) ͲͲcurrent n=237,2643,6with t smokersmoker;43.6 43 6( (mean)mean)ͲͲ smoking data(of ppastastsmokersmoker;;42.942.9 238238,371)/,) 371)/ NiN'HlhSd(NHS)dN'NursesinNurses'HealthStudyy((NHS) )andNurses' ProspectiveProspective ToTo determinedetermine relationshiprelationship betweenbetween smokingsmoking andand (mean)(mean)ͲͲnever smo smoker;ker; EverEversmo smokedked Hernan,Hernan,2001 USA 314developed MS Health Study II (NHS II)/subjects with MS diagnosis N/A Female Questionnaire MS (100) Poser(1983) Age,Age,latitude latitude,,ancestry cohortco o t MSSsrisk NHSS5;II:25Ͳ42; (Co(Conservative) se at e) withith smokingsmoking datadata prprioriorto quest q estionnaireionnaire at baselinebaseline eexc cllu dedded 34.0 (mean) ͲͲcurrent available((ofof315MS smoker;k366();()36.6(mean)ͲͲ patients)patients) pastsmoker; 34 34.0.0 (mean)(mean)ͲͲnever ne ersmo smokerker RlCllfGlPtiti'OlRoyalCollegeofGeneralPractitioners'Oral Prospective To determineassociation between use of oral n=46n=46, 000/114 Current smoking ICD MS code (340) ThoroThorogoodgood,1998 UnitedKingdomKingdom ContraceContraceptionptionstudstudyy((womenwomenrecruitedbyby N/AN/AFFemaleemale<29<29to50+ Interview MS((100)100) AAgege,pparityarity,,socialclass cohorth contraceptivesp idSikandMSrisk dldSdevelopedp MS (i)(Conservative)() providedpyid d bby GGP GPs)/MSGPs)/MScases a attrecru recruitmentitmentremove removeddf fromrom VillardͲ ProspectiveProspective Todetermineassociationbetweenuseoforal n=17,032032/63/63 OxfordFamilFamilyyPlanninPlanninggAssociationstudstudyy((womenwomen Eversmoked ICDMScode((340)340) Britainii N/A/ Femalel 25Ͳ39 Interviewi MS()(100) Age,g,pparityi y Mackintosh, cohortcohort contraceptivescontraceptives,pregnancy, pregnancypar parityityan andd MS riskrisk developeddeveloped MS recruitedrecruited fromfrom 17 clinics)/noclinics)/no MS priorprior toto studystudy (Conservative)(Conservative) fromfrom hospitalhospital adviseradviser To explore smoking and social factors in pregnant nn=250=250 Datafrom Medical Birth Registry comprised of all <25to>34313<25to>34,31.3 Eversmoked Record linkagewith CrossͲ DatafromMedicalBirthReRegistrygistryofallbirthsinNorwaNorwayy DDahl, hl 2007 subjectsbj twith ithand dwithout ith t MSand d iinvestigate ti trisk i k NNorway cases/n=371,878 371 878 bibirths th iin NNorway ffollowing ll i 12weeks k Fl/FlFemale/Female (mean)()/25t34/<25to>34, diduringpregnancy ItInterview i MS (100) NiMSNorwegianMS NNone sectional following 12 weeks gestation/pregnant nonͲMS women ffactorsactorsffororbbirthirthaandndppregnancyregnancyoutcooutcomesmes cocontrolsntrols gestatgestation/pregnantion/pregnantMSMSwomenwomen 2299.3.3((mean)mean) ((NonNonͲ Conservative)Conservative) RRegistryegistry n=480480caseswith ith smokingdata(of489 DataDatafrfromompprospectiverospectiveMMSSstudyͲͲ Spouseooffcase,case, TodkdetermineCVriskfffactorfrequencyqyinMS 18Ͳ75,,(47.3(10.9( QuestionnaireQ in CrossCrossͲ BuffaloBuffalo, NewNew casescases)/n=167)/n=167 DataData fromfrom prospectiveprospective MS studystudyͲͲ BairdBaird MS hospitalhospital staffstaff, advertisementadvertisement respondents/healthyrespondents/healthy EverEversmo smokedked RRMS (67);(67); SPMS (26);(26); Kappus, 2015 subjects and controls and relationship between CV 24/242.4/2.4 SD)/18Ͳ75,46 .4 (13 .7 interview,medical Notstated Age,sex sectional YorkYork,,USA controlswith CenterCenter/consecutive/consecutiveMSppatientsatients consecutivesubjectssubjectsrerequiredquiredtoppassassMRIscreenand ((NonNonͲ Conservative)Conservative) PPMS((7)7) riski kfactors f tand d MRImeasures SD) recordsd smoking data (of 175 achievenormal results on neurological exam controls)l)) Questionnaire (case), n=743n 743 RRMS((52);52);SPMS((9);9); CCrossͲ TTocollect ll ddatapgpertaining i ito liflifestyle y land dother h CCanadian di MSclinics, li i,y, MS SSociety iof f CCanada, d 55Ͳ 8888,,(64.6 64 6(6.2 (6 2SD)/55 )/Ͳ CkiCurrentsmokingg 2012Canadian C di PloughmanPloughman, 2014 CanadaCanadacasescases/n=8/n=8,795 ,596 2012C Canadiananadian CommunityCommunity HealthHealth SurveySurvey 35/113.5/1.1 PPMS (13);(13); PRMS (2);(2); NeurologistNeurologist NoneNone sectional healthͲrelated factors to illustrate ageing with MS general public (newspaper ads)/MS patients 75+ (NonͲConservative) Community Health controls((weighted)weighted) benibenigngn((6);6);unknown((17)17) 125 S(tl)Survey(control) nn=22=22,240,240with smokingsmoking datadata (of(of Hordaland Health StudyͲͲGeneral Hordaland CrossͲ Todetermineassociationbetweensmokingsmokingand Hordaland, 22,312312)/86)/86MS Eversmoked RiiRiise,2003 Cli/MSihdiCountyyppopulation/MS p ppatients,otherdisease N/A Nd Notstated 40Ͳ 47 QiQuestionnaire i MS (100)() SelfSlfͲreportp AgeAg sectionalsectional MS riskrisk NorwayNorway patientspatients withwith (Conservative)(Conservative) sufferers,sufferers,healthy subjects smokingg(data(of87 MS patients)

CISͲͲ>CDMS CDMS Istanbul University Medical School neurology 30.51 (8 .74 SD) ͲͲ BarkhofMRIcriteriacriteria,, RtRetrospective ti ToTdtdetermine iassociation itibtbetweensmoking kiand d n=9595CIS patients ti t /57/57 clinic/consecutiveli i / ti didiagnosed dpatients ti twith ith CIS smokers;k29.06 2906(821(8.21 MdilMedicalrecords dor AtAgeatonset, tgender, dCIStCISsymptoms, Arikanoglu, 2013 Istanbul, Turkey N/A 252.5 Current smoking CISͲͲ>CDMS 2nd attack following cohort progression from CIS toCDMS convertedto CDMS indicating MS and not other disorders,disorders, and SD)ͲͲ nonͲ smokers(age interview lesions 1stattackk absenceabsenceo offPPMS at CIS) MSclinicsatInnsbruckMedicalUniversitUniversityy n=129129CIS Prospective To determine association between smoking and neurology departments and County 14Ͳ53,29 (age range , Medical records or Poser(1983) , Onset age,gender ,CIS symptoms , Di Pauli,Pauli, 2008 Austria patients/80 N/A 2.92.9 Current smoking CISͲͲ> >CDMSCDMS cohorth progressionpg fromf CIStoCDMS Hospital/patientsp/pl/with hCISindicating d gMSanddnot mediand)ageg)atCIS) interview McDonaldld (()(2001) ) lesions,lf,interferontreatment convertedconvertedt tooCDMS otherdisorders ,and absenceof PPMS n=10371037CIS patients ti t with smoking data (of(of 1047 CIS To determine association between multiple Multiple 33centres in 17 countries/patients with CIS Current smoking as ELISA assaysͲͲ ProsProspectivepective ppatients)/atients)/ KKuhle, hl 2015 iindividual di id l ffactors tand dprogression i ffrom CIS tto EuropeanE iditiindicating MSand dnot tother thdidisorders, dand d N/AN/A221.12626Ͳ39, 3932 (median) ( di ) percotinineti i llevel l> examineriblidtblindto CISͲͲ>CDMS CDMS PoserP(1983) NoneN cohort 618converted to CCDMSDMS countries absenceofPPMSandneuromyelitisoptica 1144ng/ml clinicalinformation CDMSwithhksmokingg data available (of 623CDMS patients) ti t )

RRMSͲͲ>SPMS n=891891ppatients ti twith ith MMcDonald D ld (2001((2001, ToTo determinedetermine relationshiprelationship betweenbetween smokingsmoking andand BostonBoston, 3(based(based on 16Ͳ75 ,42 ( (mean)mean) Prospective RRMS onset (of 1465 PartnersMS Center/CDMS referrals with RRMS at 2005) (MS diagnosis)/ Healy,y,2009 MSpgprogressionoutcomes,,including gtimeto Massachusetts,, N/A/ n=1465MS (based( onn=1465MS Eversmoked QuestionnaireQ RRMSͲͲ>SPMS Age,g,sex, ,diseaseduration,,treatment cohortcohort MS patients)/72patients)/72 onset FollowFollowͲ up conversion from RRMS to SPMS USA patients) patients) develodevelopedpedSPMS (p(progression)rogression) Poser(1983) blind to nn= 179patientspatients withwith exposure datadata (MS 242.4(based on <30to ш50,36 .0 (9 .1 ProsProspectivepective Todeterminerelationshiprelationshipbetweensmokinsmokinggand RRMSonset((ofof201 GPRD((>3M>3MBritishcitizensincludedincluded)/MS)/MSppatientsatients diadiagnosis)/gnosis)/ AAgegeatinitialMSsymptoms,symptoms,sexsex,, HHernan, 2005 BitiBritain N/A n=201201MS SD)(b(based donn=201 201 EverEkdsmoked MedicalMdilrecords d RRMSͲͲ>SPMS SPMS cohort conversionconversion from RRMS to SPMS MS patients)/20 withwith RRMS at onset FollowFollowͲ upup, LublinLublin & motor symptomsymptom onset patients) MS patients) developedp SPMS Reingoldg criteria (progression) n=271271patients ti twith ith PPoser(1983) (MS To determine relationship between smoking and MS database comprised of data gathered at 25Ͳ41 ,32 (median age Prospective Groningen,Groningen, RRMS onset (out of 2.4(basedon2.4(basedon Questionnaire or diagnosis)/ Koch,h,2007 MSprogressionpg anddisability,bly,including l gconversion universityy/MSclinic/definitel/f MSpatients pwithh N/A/ atMSonsetbbasedon Eversmokedk RRMSͲͲ>SPMS MSonsetage,g,ggender cohortcohort NetherlandsNetherlands 364 MS patients)/patients)/ nn= 271 patients)patients) interviewinterview FollowFollowͲ up from a relapsingͲremitting onset to SPMS RRMS at onset n=364 patients) 117develodevelopedpedSPMS (p(progression)rogression) Poser(1983) (MS n=96ppatientswithh d)/diagnosis)/g)/ 33( (medianmedian ageat MS Retrospective To determine relationship between smoking and Vasterbotten RRMS onset (of122 National MS database from survey/incident MS 21(basedon2.1(basedon Medical records, SundstromSundstrom,,2008 NN/A/A onsetbasedonn96n=96 Eversmoked QuestionnaireRRMS ͲͲ >SPMS MSonsetaage,ge,ggenderender cohortht MSprogression irisk i k CtSdCounty,Sweden MSpatients)/ ti t )/ patientsti twith ith RRMSatttonset n=9696patients) ti t ) neurologicallilexam, patients) 26developed SPMS interview (progression)(progression) McDonald((2005)2005)((MSMS n=400400patients ti twith ith CrossͲ To determine relationship between smoking and Guilan MS Society database/CDMS patients with diagnosis)/ Roudbari,Roudbari, 2013 Iran RRMS onset/144 N/A 2.82.8 34.834.8 (9.5(9.5 SD) Eversmoked Questionnaire RRMSͲͲ> >SPMSSPMS MS onset age,age,gender gender,,relapses sectionall conversionfromf RRMStoSPMS RRMSanddSPMS Neurologistsl g SPMS patientspatients (progression) 126 APPENDIX F: Excluded Full-Text Studies, with Corresponding Reasons

1. Smoking Not Investigated

1. Kurup RK, Kurup PA. Hypothalamic digoxin and hypomagnesemia in relation to

the pathogenesis of multiple sclerosis. J Trace Elem Exp Med 2002;15:211-20.

2. Riise T, Kirkeleit J, Aarseth JH, et al. Risk of MS is not associated with exposure

to crude oil, but increases with low level of education. Mult Scler

2011;17(7):780-7.

2. Smoking Not Investigated as Independent Variable in Interaction

1. Hedstrom AK, Akerstedt T, Hillert J, Olsson T, Alfredsson L. Shift work at

young age is associated with increased risk for multiple sclerosis. Ann Neurol

2011;70:733-41.

2. Sundqvist E, Bergstrom T, Daialhosein H, et al. Cytomegalovirus seropositivity

is negatively associated with multiple sclerosis. Mult Scler 2014;20(2):165-73.

3. Sundqvist E, Sundstrom P, Linden M, et al. Epstein-Barr virus and multiple

sclerosis: Interaction with HLA. Genes Immun 2012;13(1):14-20.

3. Effects of Smoking on MS Not Reported

1. Bangsi D, Ghadirian P, Ducic S, et al. Dental amalgam and multiple sclerosis: A

case-control study in Montreal, Canada. Int J Epidemiol 1998;27(4):667-71.

2. Harpsoe MC, Basit S, Andersson M, et al. Body mass index and risk of

autoimmune diseases: A study within the Danish National Birth Cohort. Int J

Epidemiol 2014;43(3):843-55.

3. Munger KL, Chitnis T, Ascherio A. Body size and risk of MS in two cohorts of

US women. Neurology 2009;73(19):1543-50.

127 4. Riise T, Mohr DC, Munger KL, Rich-Edwards JW, Kawachi I, Ascherio A.

Stress and the risk of multiple sclerosis. Neurology 2011;76(22):1866-71.

5. Tyrpien-Golder K, Dobosz C, Damasiewicz-Bodzek A, Labuz-Roszak B,

Pierzchala K. Application of a modified high-performance thin-layer

chromatographic-densitometric technique to evaluate tobacco smoke exposure of

multiple sclerosis patients. J Planar Chromatogr 2014;27:416-9.

6. Wagner HJ, Munger KL, Ascherio A. Plasma viral load of Epstein-Barr virus and

risk of multiple sclerosis. Eur J Neurol 2004;11(12):833-4.

7. Weinstock-Guttman B, Zivadinov R, Horakova D, et al. Lipid profiles are

associated with lesion formation over 24 months in interferon-beta treated

patients following the first demyelinating event. J Neurol Neurosurg Psychiatry

2013;84(11):1186-91.

8. Weng X, Liu L, Barcellos LF, Allison JE, Herrinton LJ. Clustering of

inflammatory bowel disease with immune mediated diseases among members of

a northern California-managed care organization. Am J Gastroenterol

2007;102(7):1429-35.

9. Wesnes K, Riise T, Casetta I, et al. Body size and the risk of multiple sclerosis in

Norway and Italy: The EnvIMS study. Mult Scler 2015;21(4):388-95.

10. Zhang SM, Hernan MA, Olek MJ, Spiegelman D, Willett WC, Ascherio A.

Intakes of carotenoids, vitamin C, and vitamin E and MS risk among two large

cohorts of women. Neurology 2001;57(1):75-80.

128 4. Impossible to Isolate Effects of Smoking

1. De Jager PL, Chibnik LB, Cui J, et al. Integration of genetic risk factors into a

clinical algorithm for multiple sclerosis susceptibility: A weighted genetic risk

score. Lancet Neurol 2009;8(12):1111-9.

2. Guimond C, Lee JD, Ramagopalan SV, et al. Multiple sclerosis in the Iranian

immigrant population of BC, Canada: Prevalence and risk factors. Mult Scler

2014;20(9):1182-8.

5. Participants Matched on Smoking

1. Ragnedda G, Leoni S, Parpinel M, et al. Reduced duration of breastfeeding is

associated with a higher risk of multiple sclerosis in both Italian and Norwegian

adult males: The EnvIMS study. J Neurol 2015;262(5):1271-7.

6. Outcomes of Interest Not Investigated

1. Brown RF, Valpiani EM, Tennant CC, et al. Longitudinal assessment of anxiety,

depression, and fatigue in people with multiple sclerosis. Psychol Psychother

2009;82(Pt 1):41-56.

2. Cioncoloni D, Innocenti I, Bartalini S, et al. Individual factors enhance poor

health-related quality of life outcome in multiple sclerosis patients. Significance of

predictive determinants. J Neurol Sci 2014;345(1-2):213-9.

3. Courville CB, Maschmeyer JE, Delay CP. Effects of smoking on the acute

exacerbations of multiple sclerosis. Bull Los Angel Neuro Soc 1964;29:1-6.

4. D'Hooghe MB, Haentjens P, Nagels G, De Keyser J. Alcohol, coffee, fish,

smoking and disease progression in multiple sclerosis. Eur J Neurol

2012;19(4):616-24.

129 5. Emre M, de Decker C. Effects of cigarette smoking on motor functions in patients

with multiple sclerosis. Arch Neurol 1992;49(12):1243-7.

6. Farez MF, Fiol MP, Gaitan MI, Quintana FJ, Correale J. Sodium intake is

associated with increased disease activity in multiple sclerosis. J Neurol

Neurosurg Psychiatry 2015;86(1):26-31.

7. Gholipour T, Healy B, Baruch NF, Weiner HL, Chitnis T. Demographic and

clinical characteristics of malignant multiple sclerosis. Neurology

2011;76(23):1996-2001.

8. Jawahar R, Oh U, Eaton C, Wright N, Tindle H, Lapane KL. Association between

smoking and health outcomes in postmenopausal women living with multiple

sclerosis. Mult Scler Int 2014;2014:686045.

9. Jick SS, Li L, Falcone GJ, Vassilev ZP, Wallander MA. Epidemiology of multiple

sclerosis: Results from a large observational study in the UK. J Neurol

2015;262:2033-41.

10. Lalmohamed A, Bazelier MT, Van Staa TP, et al. Causes of death in patients with

multiple sclerosis and matched referent subjects: A population-based cohort study.

Eur J Neurol 2012;19(7):1007-14.

11. Mandia D, Ferraro OE, Nosari G, Montomoli C, Zardini E, Bergamaschi R.

Environmental factors and multiple sclerosis severity: A descriptive study. Int J

Environ Res Public Health 2014;11(6):6417-32.

12. Manouchehrinia A, Tench CR, Maxted J, Bibani RH, Britton J, Constantinescu

CS. Tobacco smoking and disability progression in multiple sclerosis: United

Kingdom cohort study. Brain 2013;136(Pt 7):2298-304.

130 13. Manouchehrinia A, Weston M, Tench CR, Britton J, Constantinescu CS. Tobacco

smoking and excess mortality in multiple sclerosis: A cohort study. J Neurol

Neurosurg Psychiatry 2014;85(10):1091-5.

14. Marrie RA, Horwitz R, Cutter G, Tyry T, Campagnolo D, Vollmer T.

Comorbidity delays diagnosis and increases disability at diagnosis in MS.

Neurology 2009;72(2):117-24.

15. Mowry EM, Waubant E, McCulloch CE, et al. Vitamin D status predicts new

brain magnetic resonance imaging activity in multiple sclerosis. Ann Neurol

2012;72(2):234-40.

16. Ozcan ME, Ince B, Bingol A, et al. Association between smoking and cognitive

impairment in multiple sclerosis. Neuropsychiatr Dis Treat 2014;10:1715-9.

17. Schwartz CE, Snook E, Quaranto B, Benedict RHB, Vollmer T. Cognitive reserve

and patient-reported outcomes in multiple sclerosis. Mult Scler 2013;19(1):87-

105.

18. Sena A, Couderc R, Ferret-Sena V, et al. Apolipoprotein E polymorphism

interacts with cigarette smoking in progression of multiple sclerosis. Eur J Neurol

2009;16(7):832-7.

19. Sena A, Couderc R, Vasconcelos JC, Ferret-Sena V, Pedrosa R. Oral

contraceptive use and clinical outcomes in patients with multiple sclerosis. J

Neurol Sci 2012;317(1-2):47-51.

20. Taylor KL, Hadgkiss EJ, Jelinek GA, et al. Lifestyle factors, demographics and

medications associated with depression risk in an international sample of people

with multiple sclerosis. BMC Psychiatry 2014;14:327.

131 21. Turner AP, Hartoonian N, Maynard C, Leipertz SL, Haselkorn JK. Smoking and

physical activity: Examining health behaviors and 15-year mortality among

individuals with multiple sclerosis. Arch Phys Med Rehabil 2015;96(3):402-9.

22. Turner AP, Kivlahan DR, Kazis LE, Haselkorn JK. Smoking among veterans with

multiple sclerosis: Prevalence correlates, quit attempts, and unmet need for

services. Arch Phys Med Rehabil 2007;88(11):1394-9.

23. Weiland TJ, Hadgkiss EJ, Jelinek GA, Pereira NG, Marck CH, van der Meer DM.

The association of alcohol consumption and smoking with quality of life,

disability and disease activity in an international sample of people with multiple

sclerosis. J Neurol Sci 2014;336(1-2):211-9.

24. Weiland TJ, Jelinek GA, Marck CH, et al. Clinically significant fatigue:

Prevalence and associated factors in an international sample of adults with

multiple sclerosis recruited via the internet. PLoS One 2015;10(2):e0115541.

25. Zivadinov R, Weinstock-Guttman B, Hashmi K, et al. Smoking is associated with

increased lesion volumes and brain atrophy in multiple sclerosis. Neurology

2009;73(7):504-10.

7. Diagnosed MS Not Investigated

1. Ponsonby AL, Lucas RM, Dear K, et al. The physical anthropometry, lifestyle

habits and blood pressure of people presenting with a first clinical demyelinating

event compared to controls: The Ausimmune study. Mult Scler

2013;19(13):1717-25.

2. van der Mei I, Lucas RM, Taylor BV, et al. Population attributable fractions and

joint effects of key risk factors for multiple sclerosis. Mult Scler 2015.

132 doi:10.1177/1352458515594040.

8. Unoriginal Research Study

1. Lauer K. Divergent risk of multiple sclerosis in two anabaptist communities in

America. Med Hypotheses 2006;67(4):969-74.

2. Pakpoor J, Goldacre R, Disanto G, Giovannoni G, Goldacre MJ. Alcohol misuse

disorders and multiple sclerosis risk. JAMA Neurology 2014;71(9):1088-9.

9. Ecological Study

1. Handel AE, Handunnetthi L, Giovannoni G, Ebers GC, Ramagopalan SV.

Genetic and environmental factors and the distribution of multiple sclerosis in

Europe. Eur J Neurol 2010;17(9):1210-4.

2. Handel AE, Jarvis L, McLaughlin R, Fries A, Ebers GC, Ramagopalan SV. The

epidemiology of multiple sclerosis in Scotland: Inferences from hospital

admissions. PLoS One 2011;6(1):e14606.

3. Palacios N, Alonso A, Bronnum-Hansen H, Ascherio A. Smoking and increased

risk of multiple sclerosis: Parallel trends in the sex ratio reinforce the evidence.

Ann Epidemiol 2011;21(7):536-42.

4. Tsai CP, Lee CT. Multiple sclerosis incidence associated with the soil lead and

arsenic concentrations in Taiwan. PLoS One 2013;8(6):e65911.

10. No Comparator Group

1. Etemadifar M, Afzali P, Tabrizi N, Hosseini SA. Pediatric multiple sclerosis with

primary progressive course--report of a retrospective cohort study in Iran.

Neuropediatrics 2013;44(3):167-70.

133 2. Marrie RA, Horwitz RI, Cutter G, Tyry T, Vollmer T. Association between

comorbidity and clinical characteristics of MS. Acta Neurol Scand

2011;124(2):135-41.

3. Staff NP, Lucchinetti CF, Keegan BM. Multiple sclerosis with predominant,

severe cognitive impairment. Arch Neurol 2009;66(9):1139-43.

11. Unhealthy Control Group

1. Brosseau L, Philippe P, Methot G, Duquette P, Haraoui B. Drug abuse as a risk

factor of multiple sclerosis: Case-control analysis and a study of heterogeneity.

Neuroepidemiology 1993;12(1):6-14.

2. Hawkes CH, Boniface D. Risk associated behavior in premorbid multiple

sclerosis: A case-control study. Mult Scler Relat Disord 2014;3(1):40-7.

3. Pekmezovic T, Drulovic J, Milenkovic M, et al. Lifestyle factors and multiple

sclerosis: A case-control study in Belgrade. Neuroepidemiology 2006;27(4):212-

4. Sternberg Z, Leung C, Sternberg D, et al. The prevalence of the classical and

non-classical cardiovascular risk factors in multiple sclerosis patients. CNS

Neurol Disord Drug Targets 2013;12(1):104-11.

12. Duplicate Data

1. Baarnhielm M, Hedstrom AK, Kockum I, et al. Sunlight is associated with

decreased multiple sclerosis risk: No interaction with human leukocyte antigen-

DRB1*15. Eur J Neurol 2012;19:955-62.

134 2. Briggs FB, Acuna BS, Shen L, et al. Adverse socioeconomic position during the

life course is associated with multiple sclerosis. J Epidemiol Community Health

2014;68(7):622-9.

3. Gianfrancesco MA, Acuna B, Shen L, et al. Obesity during childhood and

adolescence increases susceptibility to multiple sclerosis after accounting for

established genetic and environmental risk factors. Obes Res Clin Pract

2014;8:e435-47.

4. Hedstrom AK, Hillert J, Olsson T, Alfredsson L. Nicotine might have a

protective effect in the etiology of multiple sclerosis. Mult Scler

2013;19(8):1009-13.

5. Munger KL, Peeling RW, Hernan MA, et al. Infection with chlamydia

pneumoniae and risk of multiple sclerosis. Epidemiology 2003;14(2):141-7.

6. Nortvedt MW, Riise T, Maeland JG. Multiple sclerosis and lifestyle factors: The

Hordaland Health Study. Neurol Sci 2005;26(5):334-9.

7. Ueda P, Rafatnia F, Baarnhielm M, et al. Neonatal vitamin D status and risk of

multiple sclerosis. Ann Neurol 2014;76(3):338-46.

13. Comparison with Previously Published Data

1. Paul L, Weinert C. Wellness profile of midlife women with a chronic illness.

Public Health Nurs 1999;16(5):341-50.

135