AE: adverse events; BMI: body mass index; BP: blood pressure; CRF: case report forms; CTSO: clinical trial support office; DAG: data access group; DM: data manager; EGRAC: erythrocyte gluthathione reductase activation coefficient; US FNB: United States Food and Nutrition Board; FAD: flavin-adenine dinucleotide; GCP: Good Clinical Practice; HIGH: Hepcidin and Iron in Global Health; MRCG at LSHTM: Medical Research Council Unit The Gambia at London School of Hygiene and Tropical Medicine; MTHF: methylene tetrahydrofolate; MTHFR: methylene tetrahydrofolate reductase; QC: quality check; PI: principal investigator; REDCap: Research Electronic Data Capture; RNI: Recommended Nutrient Intake; SAE: severe adverse events; SEM: structural equation modelling; SSA: sub-Saharan Africa.

Hypertension is a global public health problem affecting 1 billion people worldwide and up to 46% of adults in sub-Saharan Africa (SSA) ¹ . Despite a generally very low body mass index (BMI) in the rural Kiang West district of The Gambia, the prevalence of high blood pressure (BP) (>140/90) has recently been estimated as 18.3% in adults over 18 years ² . Hypertension arises from a combination of multiple lifestyle and genetic factors (many of which probably remain unknown). Its treatment requires an individualized combination of drug therapy and lifestyle changes ^{3, 4} . The treatment of hypertension is generally lifelong and exerts a heavy economic burden on individuals and their families as well as on national economies. The identification of safe cost-effective treatment interventions without adverse side effects would be hugely advantageous, particularly in low-income settings with high prevalence of hypertension such as SSA.

Emerging evidence from clinical and genome-wide association studies (GWAS) links a functional polymorphism in the methylenetetrahydrofolate reductase ( MTHFR) gene (rs1801133), encoding the folate-metabolising enzyme MTHFR, with blood pressure (BP) and hypertension in adults ^{5, 6} . This enzyme catalyses the conversion of 5,10-methylenetetrahydrofolate (MTHF) to 5-MTHF. Variation at rs1801133 is relatively common and has 3 genotypes; homozygous “normal” wild-type CC, heterozygous CT and homozygous “variant” TT genotypes. Of these genotypes, the homozygous variant TT is strongly associated with a higher BP. The precise mechanism by which MTHFR C677T is associated with BP remains unclear.

Vitamin B2 (riboflavin) is converted to flavin-adenine dinucleotide (FAD) which acts as the essential co-factor for the MTHFR enzyme. Molecular studies demonstrate that the decreased activity of the variant enzyme is caused by reduced affinity for the redox cofactor FAD ⁷ ; hence its original name as the ‘heat-labile’ variant. Work at Ulster University has demonstrated, in three separate randomised controlled trials to date, that BP is highly responsive to intervention with riboflavin and that this response is dependent on MTHFR genotype ^{8– 10} . In all of these trials, riboflavin supplementation resulted in significant reductions in both systolic and diastolic BP in adults with the homozygous variant TT genotype, while at baseline (pre-intervention) we observed intermediate BP values in those with the heterozygous CT genotype compared to adults with CC (lower) or TT (higher) genotypes.

In the Keneba biobank ¹¹ , a total of 2114 individuals between the ages of 18 and 70 were genotyped for rs1801133 of whom 1951 (92.3%) and 163 (7.7%) had the CC and CT genotypes respectively, while none had the TT genotype. We hypothesize that the almost universal severe riboflavin deficiency seen in most Kiang West residents of The Gambia (assessed historically ¹² ) and more recently ¹³ will a) mimic the effect of the TT variant and b) exacerbate the effect of the CT genotype based upon the following argument. In vitro studies show that the TT variant is only 30% saturated with FAD compared to CC and the CT is 84% saturated ¹⁴ . Riboflavin deficiency is a major problem in Kiang West district with typical values for erythrocyte gluthathione reductase activation coefficient (EGRAC; gold-standard riboflavin biomarker) of >2.0 (compared with indicative values of <1.2, 1.2-1.4 and >1.4, defining normal, suboptimal and deficient status, respectively), indicating that EGR is only 50% saturated by FAD. By analogy, we expect these levels of severe deficiency to lower levels of MTHFR activity by ~50% in CC the genotype and a further 16% in CT variants. This is the basis for the anticipated efficacy of riboflavin supplementation.

The adult recommended nutrient intake (RNI) for riboflavin is 1.3 mg/day for men and 1.1 mg/day for women ¹⁵ . There is no known toxicity (due to a renal threshold whereby unused riboflavin is excreted rapidly in the urine) and the US FNB has noted that there have been no known adverse effects, even at intakes of 400 mg/day for at least 3 months ¹⁶ . We propose to use just 5 mg/day to replenish body stores and attempt to saturate the CT genotype enzyme as well as aligning the dose with previous trials.

The possibility that riboflavin deficiency represents a new, easily-correctible causal factor in hypertension in SSA would require further large-scale interventions if this study yields encouraging results.

We shall use regression to compare differences in BP at week 16 between the two arms of the trial after adjusting for the BP at week 0. The models are summarized below Systolic BP Week 16 = β 0 + β 1 Arm i + β 2 Systolic BP Week 0 Diastolic BP Week 16 = β 0 + β 1 Arm i + β 2 Diastolic BP Week 0 for i = (1 for riboflavin and 0 for placebo)

We would perform diagnostics to investigate normality and heretoskedasticity so as to perform a transformation as necessary.

The analyses on BP would be for systolic and diastolic separately. We shall use structural equation modelling (SEM) to predict a latent variable that could reduce the dimension from two to one variable that could reasonably summarize overall change in BP. This would be done by predicting a latent variable for both systolic and diastolic BP at week 0 and week 16 before comparing the two arms using regression as summarized in the model below. Overall BP week 16 = β 0 + β 1 Arm i + β 2 Overall BP week 0

Assessing the relationship between the systolic and diastolic BP with the overall BP would be assessed to see if the relationship is clinically plausible. If it is, then it can be used to test whether, holistically, BP at week 16 is different between the riboflavin arm and the placebo arm, adjusting for overall BP at week 0.

Missing data : Missing data less than 5% (below the 102 participants required for adequate power in the sample size calculation) would be considered acceptable to do a complete case analysis. In the event that the missing data is or above 5% we shall use Multiple Imputation to minimize bias.

Interpretation : We shall reject the null hypothesis that there is no difference between the riboflavin arm and the placebo arm at a 5% significance level. A statistically significant result between the arms where the mean BP (separated or overall) for participants that received riboflavin is lower than the mean of the participants that received placebo would mean that riboflavin successfully lowers BP 16 weeks after administration.

We shall use regression to ascertain whether there is effect modification of the intervention due to the genotype. The models are summarized below Systolic BP Wk 16 = β 0 + β 1 Systolic BP Wk 0 + β 2 Arm i + β 3 Genotype j + β 23 Arm i Genotype j Diastolic BP Wk 16 = β 0 + β 1 Diastolic BP Wk 0 + β 2 Arm i + β 3 Genotype j + β 23 Arm i Genotype j for i = (1 for riboflavin and 0 for placebo)

Similarly diagnostics would be performed to assess none of the assumptions of regression were violated and ensure that proper methods would be used to address any violations.

Interpretation : We shall reject the null hypothesis that there is no effect modification at a 5% significance level.

We shall use regression to ascertain whether there is an association between EGRAC based riboflavin status and systolic and diastolic BP at both week 0 and week 16. This is when we are considering the BPs as continuous. The models are summarized below. EGRAC Riboflavin wk 0 = β 0 + β 1 Systolic BP wk 0 EGRAC Riboflavin wk 16 = β 0 + β 1 Systolic BP wk 16 EGRAC Riboflavin wk 0 = β 0 + β 0 Diastolic BP wk 0 EGRAC Riboflavin wk 16 = β 0 + β 1 Diastolic BP wk 16

The models that use a BP cut-off of 140 mmHg for systolic and 90 mmHg for diastolic are similar to the ones above. EGRAC Riboflavin wk 0 = β 0 + β 1 Systolic BP wk 0 i EGRAC Riboflavin wk 16 = β 0 + β 1 Systolic BP wk 16 i EGRAC Riboflavin wk 0 = β 0 + β 0 Diastolic BP wk 0 j EGRAC Riboflavin wk 16 = β 0 + β 1 Diastolic BP wk 16 i for i = (1 for 140 mmHg and above and 0 for less than 140 mmHg) j = (1 for 90 mmHg and above and 0 for less than 90 mmHg)

Diagnostics would be performed to assess and correct violation of any regression assumptions as appropriate.

We shall also assess the effect modification by the MTHFR genotype using both continuous and dichotomized BP measurements. These models would also be diagnosed and violations addressed accordingly.

Interpretation : We shall reject the null hypothesis that there is no effect modification at a 5% significance level.

High BP is a public health problem, especially in SSA. The causes are not fully understood in the vast majority of cases and commonly result from a combination of lifestyle and genetic factors. The treatment of high BP is often very expensive and sometimes not affordable especially in poor communities. The development of low-cost, effective treatments would therefore be hugely beneficial.

It is possible that the severe riboflavin deficiency observed in the Kiang West district and possibly in other parts of SSA may be contributing to the rising prevalence of high BP, especially among those with a genetic predisposition for developing hypertension owing to MTHFR 677CT genotype. The study will represent the first step in helping us to understand if riboflavin supplementation helps to control blood pressure in this population group. If improvements in BP were achieved in this study and proven in large-scale intervention, it would offer a novel means of controlling hypertension that would be safer, more accessible and more cost-effective than current treatment options.

A total of 122 participants have been randomised into the study at the time of submission.