Low fish consumption is associated with a small increased risk of MS

Study design and study participants

We used the Epidemiological Investigation of Multiple Sclerosis (EIMS) and Genes and Environment in Multiple Sclerosis (GEMS), which are Swedish population-based case-control studies on genetic and environmental factors for MS. The study population is the Swedish population aged 16–70 years. During the EIMS study period, April 2005–June 2015, incident cases of MS were recruited via 42 neurology units in Swedish hospitals, including all university hospitals. Cases were diagnosed by a neurologist according to the McDonald criteria.⁹ With regard to selection of controls, we have a link with direct access to the Swedish Tax Agency. For each included case in EIMS, we request 2 controls matched by age in 5-year intervals, sex, and residential area. The controls are randomly selected among those eligible by software designed for random selection at the Tax Agency. During the GEMS study period, November 2009–November 2011, all cases registered in the Swedish MS registry¹⁰ who fulfilled the McDonald criteria⁹ were asked to participate in GEMS if they were not included in EIMS. One control for each case was randomly selected from the national population register, matched by age in 5-year intervals, sex, and residential area where the case received its diagnosis. Controls were selected for all cases at the same time, using the same procedure as in EIMS.

Standard protocol approvals, registrations, and patient consents

Both studies were approved by the Regional Ethical Review Board at Karolinska Institutet, and all participants gave their informed consent to participate.

Data collection

In both studies, information regarding environmental exposures and lifestyle factors was collected using a standardized questionnaire. The EIMS and GEMS questionnaires were similar but not identical. Completed EIMS questionnaires were obtained from 2,880 cases and 6,122 controls, with a response rate of 93% for cases and 73% for controls. Completed GEMS questionnaires were obtained from 6,156 cases and 5,408 controls, with a response rate of 82% for cases and 66% for controls. Those who could not specify their sun exposure habits or fish consumption habits were excluded (79 cases and 95 controls in EIMS and 458 cases and 245 controls in GEMS). HLA status was available for 2,053 EIMS cases and 2,878 EIMS controls and for 4,861 GEMS cases and 3,712 GEMS controls, and these were included in the present study.

Self-reported exposure information

In EIMS, participants were asked how often, on average, they have consumed lean and fatty fish during the last 5 years. In GEMS, participants were asked how often, on average, they consumed lean and fatty fish at age 20 years. Fatty fish species were defined as species with a fat content of more than 3%, such as herring, mackerel, tuna fish, salmon, and trout, whereas lean fish species were defined as species with a fat content of less than 3%, such as cod, pollock, haddock, whiting, and pike perch. Each answer alternative was reported on a 4-point scale, with the answer alternatives never/seldom, 1–3 times/months, weekly, or daily. Very few participants reported daily consumption of fish, and the last 2 answer alternatives were therefore merged into one. We constructed an index by adding the numbers together and thus acquired a value between 2 (the lowest exposure) and 6 (the highest exposure). The fish consumption index is illustrated in table 1. Low fish consumption was defined as having a value below the median among controls (<4), whereas more frequent consumption was defined as high fish consumption. In EIMS, information on sun exposure habits during the last 5 years was collected by asking 3 questions regarding ultraviolet radiation exposure.⁵ Each answer alternative was reported on a 4-point scale, and by adding the numbers together, we constructed an index ranging between 3 (the lowest exposure) and 12 (the highest exposure).

The same questions and the same answer alternatives regarding sun exposure were used in GEMS, but participants were instructed to estimate their sun exposure during summer and winter, respectively, in 10-year intervals. Sun exposure between age 10 and 29 years was considered when constructing a similar sun exposure index in GEMS, ranging between 6 (the lowest exposure) and 24 (the highest exposure). In both studies, low sun exposure was defined as having a value below the median among controls (<6 in EIMS and <12 in GEMS), whereas more frequent exposure was defined as high sun exposure.

Genotyping and measurement of vitamin D

HLA-DRB1 and HLA-A alleles were determined at 4-digit resolution. Genotyping was performed on the MS replication chip,¹¹ which is based on an Illumina exome chip to which approximately 90,000 custom markers were added with extra high density in the HLA region, and HLA alleles were then imputed with HLA*IMP:02.¹² For EIMS participants recruited between 2005 and 2009 (N = 2,462), vitamin D status was measured as levels of 25-hydroxyvitamin D using a chemiluminescent immunoassay from DiaSorin (DiaSorin AB, Sundbyberg, Sweden) and a Liaison instrument provided by DiaSorin AB with equimolar measurement of both 25-hydroxyvitamin D₂ and D₃. All samples were analyzed in a single batch. Vitamin D deficiency was defined as a value less than 50 nM/L.

Statistical analysis

Subjects with different fish consumption habits were compared with regard to MS risk by calculating ORs with 95% CIs using unconditional logistic regression models.¹³ Lean and fatty fish were also analyzed separately. The trend test for a dose-response relationship regarding fish consumption and risk of MS was performed by using a continuous variable for fish consumption (the fish consumption index) in a logistic regression model. Trend tests for dose-response relationships were also performed for lean and fatty fish separately. The analysis of fish consumption and MS risk was stratified by sex.

Causal mediation analysis was performed to asses to what extent the relationship between consumption of fatty fish and MS risk was mediated by vitamin D status. The exposure variable was low consumption of fatty fish, the mediator was vitamin D deficiency, and the outcome MS. The causal effects were estimated on the OR scale, and the CIs were calculated using the delta method.¹⁴

Potential biologic interactions were assessed on an additive scale by calculating the attributable proportion due to interaction with 95% CI. To elucidate the relationship between fish consumption, sun exposure, and DRB1*15:01 status, the analysis of interaction between low fish consumption and carriage of DRB1*15:01 was stratified by sun exposure habits. We also assessed potential interactions between low sun exposure and DRB1*15:01, stratified by fish consumption habits, and between low sun exposure and low fish consumption, stratified by DRB1*15:01 status. Similar analyses were performed to study the relationship between fish consumption, sun exposure, and A*02:01 status.

All analyses were adjusted for age, residential area, ancestry, and the following MS-associated HLA alleles that have been associated with MS risk independently of DRB1*15:01 status¹: DRB1*03:01, DRB1*13:03, DRB1*08:01, B*44:02, B*38:01, B*55:01, DQA1*01:01, DQB1*03:02, and homozygote correction for DRB1*15:01, DRB1*03:01, and A*02:01. Ancestry was dichotomized into Nordic vs non-Nordic origin. A participant who was born in any of the Nordic countries, whose parents had not immigrated from outside the Nordic countries, was classified as Nordic. When appropriate, the analyses were also adjusted for sex, sun exposure habits, fish consumption habits, DRB1*15:01, and A*02:01 status.

Adjustments were also made for a number of potential confounding variables that only had minor influence on the results. These factors were not kept in the final analyses and included educational level, intake of vitamin supplements, adolescent body mass index (BMI), a history of infectious mononucleosis, smoking, passive smoking, and alcohol consumption. Educational level was categorized into no postsecondary education, postsecondary education without a university degree, or university degree. Intake of vitamin supplements was dichotomized into regular intake of multivitamins or vitamin D supplements or not. Adolescent BMI was calculated by dividing self-reported weight in kilograms by self-reported height in meters squared and dichotomized into BMI <25 or BMI ≥25 kg/m². A history of mononucleosis was dichotomized into yes or no. Smoking and passive smoking were considered before the onset of disease among cases and during the same period among the corresponding controls. Smoking was dichotomized into current, past, or never smokers. Passive smoking was dichotomized into ever or never exposed. Alcohol consumption was categorized into low, moderate, or high consumption based on the number of drinks per week.¹⁵

We performed 2 subanalyses based on EIMS to be able to adjust the analyses for dietary habits and salt intake. Participants with disease onset during the last 5 years before study inclusion were included in the first subanalysis (2,209 cases and 4,610 controls). Based on a question regarding dietary habits 5 years ago, participants could choose between the following answer alternatives (1) normal diet/mixed diet, (2) vegetarian food, (3) vegetarian food including fish and shellfish, (4) vegan diet, (5) Mediterranean diet, (6) GI diet, and (7) other diet. Dietary habits were categorized based on the answer alternatives, using normal diet/mixed diet as the reference group. In the second subanalysis, we adjusted for salt intake. In November 2013, complementary questions were sent to all participants who had answered the standardized questionnaire during April 2005 to March 2013. The complementary questions were answered by 1,762 cases (82%) and 2,975 controls (66%). Among other questions, participants were asked to report their salt intake during different age periods. The questions were as follows: (1) how often do you eat at restaurants, including takeaway? (2) how often do you eat prepared meals or dishes that are bought at the supermarket, and (3) how often do you use extra salt on your food? The answer alternatives were less than 1 time/month, 1–3 times/month, 1–2 times/week, 3–6 times/week, daily, and several times/day. Each question was categorized based on the answer alternatives, using less than 1 time/month as the reference group. All analyses were conducted using Statistical Analysis System (SAS) version 9.4.

Data availability

Anonymized data will be shared by request from any qualified investigator who wants to analyze questions that are related to the published article.