This prevalent case-control study was designed and reported using Strengthening the Reporting of Observational Studies in Epidemiology—Nutritional Epidemiology (STROBE-nut) guideline, which refers to a simple reporting guideline checklist used to improve the quality of the research design (29). This study included 541 patients diagnosed with MS by a multidisciplinary neurology team in Jordan using the International Classification of Diseases version 10 (ICD-10) and following the revised McDonald criteria in 2017 for MS Diagnosis (30). Prior to recruitment of healthy controls, the clinical signs and characteristics of MS and other severe medical comorbidities were clarified to the participants. This verification was done to avoid misidentifying and misclassifying of healthy participants as cases participants (19). Healthy controls (not from MS family) without self-report medical records or symptoms of MS (n = 607) matched on sex and age (average 42 years) were recruited from the general population from different areas of Amman. Simple random sampling was used to select the MS cases from the study site’s case registry located in Amman, Jordan. The recruitment process took place in the out-patient department, between the months of March and June in the year 2021. Data collected reflect a duration of illness that varied from 1 to 3 years.

Outpatients of both sexes diagnosed with MS, aged between 20 and 60 years, able to provide consent, and express willingness to participate in the study were eligible to participate in the study. Those patients having severe medical comorbidities (cancer, motor neuron diseases, chronic kidney disease, psychiatric disorders), pregnant women, and those participating in other experimental protocols were excluded.

Based on previous similar research (8), we estimated that a sample size of 100 patients would be required to answer the research hypothesis for our prevalent case-control analysis. Patients with MS are two times (odds ratio = 2.0) more likely to be on the extreme end of the pro-inflammatory diet (20). Therefore, the sample size calculator included the following assumptions α = 5%, β = 20% to achieve 80% power. We decided to include about 500 patients per group to increase the power further.

The Research Ethics Committee, University of Petra, Jordan (UOP/REC: Q1R/11/2020) granted ethical approval for the study. Before any data were collected, all participants signed a written informed consent form. The study adhered with the Declaration of Helsinki ethical principles for medical research involving human subjects.

The following assessments were undertaken during the clinical interview: demographics, and anthropometrics measurements (weight and height) using Seca digital stadiometer, which measures body weight with 0.1 Kg accuracy and height to the nearest ± 0.1 cm (31). Body mass index (BMI) was calculated based on measured weight and height by dividing body weight in kg by the squared height in meters. The BMI of subjects were then classified as underweight (BMI < 18.5 kg/m²), normal (BMI < 24.9 kg/m²; BMI ≥ 25.0), overweight (25), or obese <29.9; BMI ≥ 30.0, respectively by the World Health Organization (32).

The participants received one study questionnaire survey that consisted of the detailed dietary habits and other lifestyle aspects. The dietary intake was assessed using a validated Arabic food frequency questionnaire (FFQ) with Cronbach alpha and McDonald omega of >0.9 (33, 34). The FFQ is described in detail in another publication (33–35). Briefly, it includes thirty food items and eight beverages, the FFQ question asked about the average frequency of a certain food item throughout the past month as well as standard serving. The collected responses of the consumed frequency of a specific serving size was standardized using food images to assess a standard unit for portions.

Dummy models were utilized to determine a standard unit for portions to standardize data collection further. The FFQ, which measured food intake over the previous 4 weeks, was divided into five major food groups, which were further divided into 32 categories. According to a recent systematic review and meta-analysis, using FFQs (typically from the previous month) is a reliable method in nutritional epidemiological studies (36).

ESHA’s Food Processor^® nutrition analysis software was utilized to analyze dietary of the FFQ, and Dietary Inflammatory Index (DII^®) scores were calculated from these data. A total of 32 food parameters, of the possible 45, were used to calculate the DII score. These are: Energy (kcal); Protein (g); Carbohydrate (g); Fiber (g); Cholesterol (mg); Total Fat (g); Saturated Fat (g); Monounsaturated Fat (g); Polyunsaturated Fat (g); Calcium (mg); Iron (mg); Magnesium (mg); Sodium (mg); Phosphorous (mg); Potassium (mg); Zinc (mg); Niacin B3 (mg); Thiamin B1 (mg); Riboflavin B2 (mg); Vitamin A (RE); Pyridoxine B6 (mg); Cobalamin B12 (μg); Vitamin C (mg); Omega-3 (g); Omega-6 (g); Vitamin D (IU); Vitamin E (mg); Folic acid (μg); Selenium (μg); Caffeine (mg); β-carotene (μg); Alcohol (g).

An in-depth description of the DII (37) and the energy-adjusted DII (E-DII™) can be found elsewhere (38). Briefly, the DII scoring algorithm was based on a careful review of the literature through which 1943 articles identified 45 food parameters (i.e., macronutrients including specific categories of fatty acids, carbohydrate, and proteins; macronutrients including vitamins and minerals; flavonoids; and whole food items including herbs and spices) as having sufficiently robust literature in relation to six inflammatory biomarkers—i.e., interleukins (IL)-1β, –4, –6, –10, tumor-necrosis factor-alpha (TNFα), and C-reactive protein (CRP). As noted, self-report values for 32 of these food parameters were available from the FFQ. These were translated into z-scores using a global comparative database consisting of data from 11 countries by subtracting from the individual’s self-report value the mean of the global database then dividing by the standard deviation. These scores were then converted to proportions and centered on zero by doubling each and subtracting one. These centered proportions were then multiplied by their respective coefficients (overall food parameter-specific inflammatory effect scores) to obtain DII scores for each food parameter. These were summed to obtain the overall DII score. E-DII scores were calculated using the density approach by calculating DII per 1,000 kcal consumption. This employed the same procedure for scoring, but relies on an energy-adjusted global comparison database (38). These DII and E-DII scores have a potential range from approximately –9 to + 8; i.e., from minimally to maximally pro-inflammatory, respectively. The DII and E-DII are scored similarly and scaled identically; so, the scores are comparable across studies. For the E-DII, energy was in the denominator; so, 31 parameters were used for computation. The decision to use the DII was based on model goodness of fit/explanatory ability.

Before beginning the analyses, Shapiro–Wilk test was utilized to visualize the data for tested normality and formally. Means and standard deviations are reported for continuous variables, and counts and percentages are reported for categorical variables. The independent sample t-test was used for continuous variables and the Chi-square test was used for categorical variables to compare the two independent groups. With age, sex, body mass index, and smoking status as covariates, multinomial logistic regression analysis was performed on DII as the outcome variable and group assignments of the participants according to DII quartiles as the predictor. All data analysis was done using STATA 17.0 software and, as noted, statistical significance was set at α = 0.05, two-sided. We used three measures of model fit for regression analyses to determine if our data better fit DII or E-DII. The fit indices included R-squared (R²), Akaike Information Criterion (AIC); and Bayesian Information Criterion (BIC) (39). To explain further, when performing regression analysis, it is important to assess which model best fits the data. To determine if the data better fits the DII or E-DII model, three different measures of model fit must be used. R² gives the proportion of variation in the response variable which is explained by the model. The higher the R², the better the model fits the data. By taking into account these three measures of model fit, we can reasonably assess whether the data fits the DII or E-DII model better. The results provide a better understanding of the data and help inform further analysis. This can help to ensure that the best fitting model is utilized and the results are as accurate as possible.

Descriptive data for both cases and controls are shown in Table 1. MS is a more common disease in females compared to males as is reflected in the lower percentage of males among MS cases and sex-matched controls. There also was no significant difference between the cases and controls in terms of age, the other variable on which we matched. However, large and statistically significant differences were observed in the anthropometric measurements. Cases were more likely to be obese compared to controls (21.07 vs. 10.21%). Cases also were more likely to be current smokers compared to controls (24.95 vs. 16.31%). Relapsing remitting multiple sclerosis (RRMS) represents most (≈85%) of the studied cases, while the prevalence of primary progressive multiple sclerosis (PPMS) and secondary progressive multiple sclerosis (SPMS) was 15%.

The results in Table 2 are in line with the predicted direction between cases and control participants, except for the lack of effect of carbohydrate, fiber, and selenium. The DII scores are shown in Table 2. The decision to use the DII instead of the E-DII score was based on model goodness of fit using the AIC = 1,398, BIC = 1,427, R2 = 0.22 for DII and AIC = 1,499, BIC = 1,528, R2 = 0.11 for E-DII. The ORs and 95% CIs for MS status by quartiles of DII are shown in Table 3. Participants in the most pro-inflammatory group (Quartile 4) were about ten times more likely to have MS than participants in the most anti-inflammatory group (Quartile 1) [OR 10.17 (95% CI 6.88; 15.04); P-trend = 0.001] when DII was expressed as a continuous variable.

The current results from a large population-based prevalent case-control study in Jordan add to the existing MS literature on the non-genetic factors associated with MS prevalence and progression. Moreover, the findings of this study support a link between diet’s overall inflammatory potential and the odds of MS. Furthermore, the findings are in line with the use of DII as a predictor of MS disease. Such promising findings highlight the critical importance of educating MS patients and raising their awareness about food group choices and eating habits in order to improve their health and slow the progression of MS symptoms.

Despite its strengths, this study has some limitations. These includes variations in the duration of illness in MS cases, for which we used simple random sampling. We did limit recruitment only to prevalent cases and disease duration varied from 1 to 3 years, which might affect individuals’ appetites and dietary requirements. Furthermore, the results of this study are limited to a large population-based prevalent case-control study in Jordan. Therefore, questions remain regarding the ability to generalize findings to the MS patient population in Arab countries. Hence, a larger MS patient population and MS subtype shall be included in the analysis from all Arab countries. Additionally, this prevalent case-control study is subject to reporting biases that result from individuals’ knowledge of disease and the effect of disease on factors that, in turn, might affect diet. In addition, the FFQ used in this study could not capture all DII components; hence, more comprehensive FFQ is required.

Although healthy controls without MS matched on sex and age were recruited from the general population, differences were observed with respect to smoking and obesity. This could be a limitation of this population-based prevalent case-control study design. Though we controlled a variety of covariates including age, sex, body mass index, and smoking status, residual confounding still may exist. Moreover, physical activity is an important confounder and needs further future study.