Of 204 screened individuals, 114 were not eligible and six withdrew for other reasons. The remaining 84 participants were randomised (ESM Fig. 1) and constituted the safety and full analysis set (Table 1). At baseline, mean liver PDFF in all participants was 18%, HbA1c 58 mmol/mol (7.4%), C-peptide level 0.98 nmol/l and fasting plasma glucose 9.4 mmol/l. Fewer than 10% of participants reported diabetes-related complications. Most individuals were treated with metformin or a sulfonylurea (82% and 18%, respectively) alone or in combination, and 14% were drug naive. No change in this medication occurred during the study. Participants were randomised to four groups: placebo (n = 21), OM-3CA monotherapy (n = 20), dapagliflozin monotherapy (n = 21) and combined OM-3CA and dapagliflozin therapy (n = 22). Compliance was high (97%) and similar in each treatment group. In total, 75 participants (89%) completed the study (ESM Fig. 1).
Liver fat content
Dapagliflozin and OM-3CA, on their own or in combination, significantly reduced liver PDFF from baseline to 12 weeks (Table 2 and Fig. 1). The combination treatment (relative change, −21%, adjusted p < 0.05), but not dapagliflozin (−13%) or OM-3CA (−15%) as monotherapies, significantly lowered liver PDFF compared with placebo (−3%). Total liver fat volume calculated from total liver volume and PDFF changed similarly to liver PDFF (Table 2 and Fig. 1), suggesting that the change in PDFF was not secondary to changes in hepatic fat-free volume.
The baseline NAFLD fibrosis score was calculated . Eight participants (n = 2 in OM-3CA group, n = 4 in the dapagliflozin group and n = 2 in the combination group) had an NAFLD fibrosis score >0.675, indicating significant fibrosis. The NAFLD fibrosis score did not interact significantly with the effect of treatment on liver PDFF.
The I148M genetic variant of PNPLA3 (rs738409 C>G) has been reported to increase liver fat content and risk of developing NASH . Therefore, the participants with the most common genotype C/C (n = 47) were compared with those with the C/G (n = 30) and G/G genotypes (n = 3). The proportions of the C/C and C/G + G/G genotypes did not differ significantly between the treatment groups. Baseline liver PDFF was numerically lower in the C/C group (median [range]: 17.4% [8.3–34.7%]) than in the C/G + G/G group (20.0% [6.1–48.5%]; p = 0.19). There was a significant interaction between PNPLA3 genotype (C/C vs C/G + G/G) and treatment response on liver PDFF across the active treatment arms (p = 0.03). In the combination treatment group, the relative reduction in liver PDFF was numerically larger for the C/G + G/G genotype (relative change: −25.4% [−27.3 to −19.0%]) than for the C/C genotype (−16.1% [−20.5 to −11.6%]), and this was significantly different (p < 0.01) from that seen with dapagliflozin alone (C/G + G/G: 7.0% [−2.2 to 11.3%]; C/C: −22.0% [−26.8 to −19.2%]; p < 0.01). In the OM-3CA treatment group, the change in liver PDFF was numerically smaller for the C/G + G/G genotype (−12.6% [−15.9 to −4.3%]) than for the C/C genotype (−18.6% [−20.2 to −15.6%]), but this was not significantly different from other treatment groups.
Anthropometrics and abdominal adipose tissue volumes
Participants using dapagliflozin alone and in combination with OM-3CA showed reduced body weight and waist circumference, whereas those using OM-3CA alone showed no change compared with the placebo group (Table 2). Abdominal subcutaneous and visceral adipose tissue volumes decreased in the two groups treated with dapagliflozin. There was a significant interaction between the baseline subcutaneous fat volume and treatment response in the active treatment arms on liver PDFF (p = 0.017), which reached statistical significance in the group treated with both dapagliflozin and OM-3CA vs the group treated with dapagliflozin alone (p = 0.006).
In participants using dapagliflozin, HbA1c decreased from baseline, but this effect was significant only in individuals using dapagliflozin monotherapy when compared with the placebo group (Table 3). In participants using dapagliflozin, fasting and 2 h plasma glucose concentrations decreased, while OM-3CA treatment had no effect (Table 3). Fasting insulin levels were reduced from baseline in all participants using dapagliflozin, but the effect was not significant vs the placebo group. Dapagliflozin treatment significantly improved the insulin sensitivity index measured using HOMA-IR, while OM-3CA treatment had no effect. Plasma levels of NEFA and the insulin sensitivity index for NEFA/lipolysis  were not affected by any treatment. The changes in 2 h plasma glucose and insulin levels in the dapagliflozin monotherapy group correlated with the changes found in liver PDFF (ρ = 0.55, p = 0.02 and ρ = 0.62, p = 0.005, respectively; ESM Fig. 2).
Plasma fatty acid composition and lipoprotein levels
Concentrations of the fatty acids DHA and EPA were measured in total plasma and in the cholesteryl ester and phospholipid fractions (Table 4 and ESM Tables 1–4). DHA levels increased by 20–40%, while EPA levels increased about threefold in the different lipid fractions in participants using OM-3CA. Baseline levels of DHA or EPA did not interact significantly with the effect of treatment on liver PDFF, and there were no significant correlations between changes in DHA or EPA and changes in liver PDFF.
Changes in fatty acid composition in the cholesteryl ester and phospholipid fractions are shown in ESM Tables 2 and 3. OM-3CA treatment resulted in small and inconsistent changes in saturated fatty acids in the cholesteryl ester and phospholipid fractions. Levels of monounsaturated fatty acids (16:1 n-7, 18:1 n-9) and several n-6 fatty acids (18:2, 18:3, 20:3) decreased after OM-3CA treatment, while 18:3 n-3 and 20:4 n-6 levels did not change. Dapagliflozin treatment had no or inconsistent effects on fatty acid composition in the cholesteryl ester and phospholipid fractions (ESM Tables 2 and 3).
OM-3CA treatment increased estimated δ-5 desaturase activity and decreased δ-6 desaturase and stearoyl-CoA desaturase-1 (SCD-1) activity indices, while dapagliflozin treatment had no significant effect on these activities (Table 4 and ESM Table 4). The change in liver PDFF was significantly associated with change in SCD-1 index in the dapagliflozin group (ρ = 0.54, p = 0.02; ESM Fig. 2), but not in the other groups. Elongase activity index (cholesteryl ester 18:0 cholesteryl ester 16:0) was not significantly influenced by treatment (data not shown). Total, LDL- and HDL-cholesterol as well as triacylglycerol levels were not significantly changed by any treatment vs placebo (ESM Table 5). Apolipoprotein C3 levels increased following treatment with dapagliflozin, while no effect was seen in the OM-3CA groups.
Both dapagliflozin groups had increased β-hydroxybutyrate levels numerically, but not significantly vs the placebo group (ESM Table 5). No significant correlation between changes in β-hydroxybutyrate and liver PDFF was observed. Dapagliflozin treatment increased butyrylcarnitine levels, while there was no effect of the combination treatment on plasma levels of the acylcarnitines vs placebo (ESM Table 5).
Hepatocyte injury, oxidative stress and inflammation biomarkers and adipokines
Dapagliflozin monotherapy reduced levels of all measured hepatocyte injury biomarkers, including aspartate aminotransferase, alanine aminotransferase, γ-glutamyl transferase (γ-GT), CK 18-M30 and CK 18-M65 (Fig. 2, ESM Table 6). There was no significant effect of the OM-3CA monotherapy or the combination therapy on any of the hepatocyte injury biomarkers. Changes in liver PDFF correlated significantly with changes in γ-GT (ρ = 0.53, p = 0.02), but not with the other hepatocyte injury biomarkers in the dapagliflozin group. Uric acid levels were significantly reduced in the dapagliflozin and combination groups, but not with OM-3CA, compared with placebo (ESM Table 6).
Oxidative stress biomarkers related to non-enzymatic oxidation of unsaturated fatty acids were measured in plasma and urine (ESM Table 6). The n-3 fatty acid-derived oxidative stress biomarker 2-hydroxyhexenal was significantly increased by the OM-3CA and combination treatments. Levels of 8-iso-prostaglandin F2-α (8-iso-PGF2-α) were numerically lower in the OM-3CA group.
Plasma osteopontin levels increased significantly following OM-3CA + dapagliflozin, but not monotherapies, vs placebo. OM-3CA + dapagliflozin dual therapy had no significant effects on C-reactive protein, adiponectin and leptin levels (ESM Table 7). FGF21 levels decreased significantly in the dapagliflozin compared with the placebo group, but OM-3CA or combination treatments had no significant effect (ESM Table 7 and ESM Fig. 2). There was no significant association between changes in FGF21 and changes in liver PDFF, acylcarnitines or β-hydroxybutyrate in active treatment groups.
Adverse events and safety
All active treatment groups had similar total percentages of adverse event reporting (70.0–77.3%), which were higher than in the placebo group (47.6%). There were no new or unexpected adverse events compared with previous studies with these treatments. More participants reported adverse events when using dapagliflozin and OM-3CA (n = 15, 68.2%) than when using dapagliflozin monotherapy (n = 7, 33.3%), OM-3CA monotherapy (n = 8, 40%) or placebo (n = 6, 28.6%). All adverse events were mild or moderate in intensity, except two serious adverse events judged by investigators as unlikely to be caused by study treatments. There were no significant changes in serum creatinine levels in any of the treatment groups.