Table 1. Characteristics of included studies.
| Study | Country | Study design and population | Years of follow-up | Diagnosis of NAFLD at cohort inclusion | Metabolic RFs studied | Study outcomes of interest and number of events | Adjustments of interest considered | Adjusted HRs for liver events with 95% CIs and p-values | Risk of bias |
|---|---|---|---|---|---|---|---|---|---|
| Alexander 2019 [20] | UK, Netherlands, Italy, Spain | Retrospective data linkage cohort analysis, 18 million, 136,703 with NAFLD, mean age 55.8 years, 52% M | 3.3 | Yes (coding) | T2DM, BP, obesity | Cirrhosis/complications, 7,375 events | Alcohol, other metabolic RFs | *BMI > 30: 1.03 (1.03–1.04), p < 0.001; T2DM: 2.86 (2.71–3.02), p < 0.001; high BP: 1.06 (1.00–1.12), p = 0.03 | Low |
| Andreasson 2017 [21] | Sweden (Malmo cohort) | Prospective population cohort, 27,617, mean age 58.1 years, 38.8% M | 19.8 | No (exclusion of other causes of LD) | Obesity | Composite non-fatal and fatal LD, 505 events | Alcohol | *BMI > 30: 1.52 (1.17–1.98), p = 0.002; increased WC: women: 1.75 (1.32–2.33), p < 0.001, men: 1.69 (1.28–2.23), p < 0.001; increased WHR: women: 1.68 (1.36–2.07), p < 0.001, men: 1.78 (1.41–2.25), p < 0.001 | Low |
| Björkström 2019 [22] | Sweden | Retrospective data linkage cohort analysis, 2.5 million, 406,770 with T2DM, mean age 64.7 years, 53.8% M | 7.7 | No (exclusion of other causes of LD) | T2DM | Composite non-fatal and fatal LD, 16,711 events | Unclear—high alcohol risk excluded at baseline | T2DM: 2.28 (2.21–2.36), p < 0.001 | Low |
| El-Serag 2004 [23] | US | Retrospective data linkage cohort analysis, 173,643 with diabetes, 650,620 without diabetes, age > 20 years, 98% M (veterans) | 10 | No (exclusion of other causes of LD) | T2DM | Composite non-fatal and fatal LD, 7,799 events | Alcohol | T2DM: 2.15 (2.00–2.31), p < 0.001 | Mod |
| Goh 2017 [24] | Singapore (Singapore Chinese Health Study) | Prospective population cohort, 63,247, age 45–74 years, 50% M | 16.9 | No (exclusion of other causes of LD) | T2DM, obesity | LD mortality, 133 events | Alcohol | T2DM: 2.6 (1.73–3.89), p < 0.001; BMI > 30: 1.36 (0.86–2.17), p = 0.19 | Low |
| Golabi 2018 [25] | US (NHANES III) | Prospective population cohort, 3,613, median age 43 years, 50% M | 19 | Yes (ultrasound) | Metabolic syndrome | LD mortality, 22 events | Alcohol | 1 MS RF: 26.35 (2.46–282.72), p = 0.007; 2 MS RF: 16.95 (1.59–180.91), p = 0.019; 3 MS RF: 1.98 (0.11–34.38), p = 0.64; 4 MS RF: 4.57 (0.32–64.88), p = 0.26 | Mod |
| Hagström 2016$ [26] | Sweden | Prospective population cohort, 44,248, age 18–20 years, 100% M (army conscripts) | 37.8 | No (exclusion of other causes of LD) | Obesity | Composite non-fatal and fatal LD, 393 events | Alcohol, BP | BMI > 30: 1.59 (0.64–3.95), p = 0.32 | Mod |
| Hagström 2018 [27] | Sweden | Prospective population cohort, 1,220,2161, age 17–19 years, 100% M (army conscripts) | 28.5 | No (exclusion of other causes of LD) | Obesity, T2DM | Composite non-fatal and fatal LD, 5,281 events | High alcohol risk excluded at baseline, obesity, BP | T2DM: 3.49 (3.01–4.03), p < 0.001 | Low |
| Hagström 2019 [28] | Sweden | Retrospective data linkage cohort analysis, 1,185,733, mean age 28.6 years, 100% F (antenatal) | 13.8 | No (sensitivity analysis to exclude alcohol diagnoses) | Obesity, T2DM | Composite non-fatal and fatal LD, 852 events | Obesity, T2DM | BMI > 30: 1.76 (1.27–2.46), p = 0.001; T2DM: 4.30 (3.23–5.72), p < 0.001 | Low |
| Ioannou 2003 [29] | US (NHANES I) | Prospective population cohort, 11,465, age 25–74 years, 50% M | 13 | No (exclusion of cirrhosis from other causes of LD) | Obesity | Composite non-fatal and fatal LD, 89 events | T2DM, cholesterol, alcohol | BMI > 30: 1.65 (0.9–3.1), p = 0.11 | Low |
| Ioannou 2005 [30] | US (NHANES I) | Prospective population cohort, 11,434, age 25–74 years, 50% M | 13 | No (exclusion of cirrhosis from other causes of LD) | Obesity (central) | Composite non-fatal and fatal LD, 88 events | Alcohol | BMI > 30 and subscapular-to-triceps skinfold thickness ratio: high: 2.2 (1.1–4.6), p = 0.026, low: 0.8 (0.2–2.8), p = 0.75 | Low |
| Kanwal 2019 [31] | US | Retrospective data linkage cohort analysis, 271,906, mean age 55.5 years, 94.3% M | 9 | Yes (abnormal blood tests) (exclusion of other causes of LD) | All metabolic risk factors | Cirrhosis, 22,794 events | Alcohol risk excluded at baseline and throughout follow-up period, other metabolic RFs | BMI > 30: 1.09 (1.06–1.13), p < 0.001; T2DM: 1.31 (1.27–1.34), p < 0.001; high BP: 1.59 (1.51–1.69), p < 0.001; dyslipidaemia (composite): 1.23 (1.19–1.28), p < 0.001; 2 MS RF: 1.33 (1.26–1.40), p < 0.001; 3 MS RF: 1.61 (1.53–1.69), p < 0.001; 4 MS RF: 2.03 (1.93–2.13), p < 0.001 | Low |
| Liu 2010 [32] | UK (Million Women Study) | Prospective population cohort, 1,230,662, mean age 56 years, 100% F | 6.2 | No (exclusion of other causes of LD) | Obesity | Composite non-fatal and fatal LD, 1,811 events | Alcohol, BMI, T2DM | BMI > 30: 1.49 (1.33–1.68), p < 0.001; T2DM: 4.29 (2.74–6.73), p < 0.001 | Low |
| Nderitu 2017 [33] | Sweden (AMORIS cohort) | Prospective population cohort, 509,436, mean age 44 years, 53.4% M | 20 | No (exclusion of other causes of LD) | All metabolic risk factors | Cirrhosis/complications, 2,775 events | Other metabolic RFs | low HDL: 1.28 (1.04–1.59), p = 0.020; high triglycerides: 1.30 (0.99–1.72), p = 0.059; BMI > 30: 1.38 (0.93–2.04), p = 0.11; T2DM: 2.00 (1.19–3.38), p = 0.009 | Mod |
| Otgonsuren 2013 [34] | US (NHANES III) | Prospective population cohort, 10,565, age 20–50 years, 45% M | 13.8 | Yes (ultrasound) (exclusion of other causes of LD) | Obesity | LD mortality, 26 events | Alcohol, BP, T2DM | BMI > 30: 1.06 (0.96–1.16), p = 0.25; WC: 1.02 (0.98–1.07), p = 0.332; WHR > 0.8: 83.51 (2.03–3,434.26), p = 0.02 | Low |
| Pang 2018 [35] | China (China Kadoorie Biobank) | Prospective population cohort, 503,993, mean age 51.5 years, 41% M | 10 | No (exclusion of other causes of LD) | T2DM | Cirrhosis/complications, 2,082 events | Alcohol, BMI | T2DM: 1.78 (1.45–2.18), p < 0.001 | Mod |
| Porepa 2010 [36] | Canada | Retrospective data linkage cohort analysis, 2,497,777, mean age 55.3 years, 56.3% M | 6.4 | No (exclusion of other causes of LD) | T2DM, BP, obesity | Cirrhosis/complications, 8,365 events | BP, lipids, obesity, T2DM | T2DM: 1.77 (1.68–1.86), p < 0.001; high BP: 1.23 (1.14–1.31), p < 0.001; BMI > 30: 1.16 (1.01–1.33), p = 0.03 | Low |
| Schult 2011 [37] | Sweden (Gothenberg survey) | Prospective population cohort, 855, mean age 50 years,100% M | 40 | No (exclusion of other causes of LD) | All metabolic risk factors | Composite non-fatal and fatal LD, 14 events | Alcohol | BMI > 30: 1.27 (1.09–1.48), p = 0.002; triglycerides: 1.99 (1.35–2.96), p = 0.001; other HRs not presented | Mod |
| Schult 2018 [38] | Sweden (Gothenberg survey) | Prospective population cohort, 1,462, age 38–60 years, 100% F | 42 | No (exclusion of other causes of LD) | Obesity (central) | Composite non-fatal and fatal LD, 11 events | Alcohol, BP | WHR > 0.8: 5.82 (1.59–21.4), p = 0.008 | Mod |
| Simeone 2017 [39] | US | Retrospective data linkage cohort analysis, 18,754, age > 18 years, 38.5% M | 2.3 | Yes (coding) (exclusion of other causes of LD) | T2DM | Composite non-fatal and fatal LD, 5,645 events (any disease progression) | Unclear | T2DM: 2.0 (no CI given) | High |
| Stepanova 2010 [40] | US (NHANES III) | Prospective population cohort, 991, age > 17 years, 47.5% M | 13.3 | Yes (abnormal blood tests) (exclusion of other causes of LD) | All metabolic risk factors | LD mortality, 117 events | Alcohol, other metabolic RFs | T2DM: 1.05 (1–1.65), p < 0.05@; high cholesterol: 0.37 (0.06–2.15), p = 0.284; high BP: 0.07 (0.01–0.3x), p = 0.007; BMI > 30: 11.19 (2.43–51.56), p = 0.002; MS: 12.08 (1.10–132.22), p = 0.042 | Mod |
| Younossi 2013 [41] | US (NHANES III) | Prospective population cohort, 1,448, age > 18 years, 64% M | 16 | Yes (ultrasound) (exclusion of other causes of LD) | Metabolic syndrome, obesity | LD mortality, 10 events | Metabolic RFs, alcohol | BMI > 30: 1.12 (1.03–1.21), p = 0.008; MS: 294.24 (118.74–729.14), p < 0.001 | Low |
BMI units are kg/m2.
*Adjusted HR for whole cohort using a fixed-effects meta-analysis to get the combined HR from the 2 presented HRs for subgroups (coded versus uncoded combined in Alexander et al.; men and women combined in Andreasson et al.).
$Sub-cohort of the 2018 Hagström study but with additional data on alcohol consumption.
@CI and p-value as given in the paper presented here—different from the calculated CI used in meta-analysis using the HR and sample size (Fig 2). This difference is due to CI asymmetry in the published figures and inability to reproduce these figures on log transformation. Authors contacted to confirm data—no response.
BMI, body mass index; BP, blood pressure; CI, confidence interval; F, female; HDL, high-density lipoprotein; HR, hazard ratio; LD, liver disease; M, male; mod, moderate; MS, metabolic syndrome; NAFLD, non-alcoholic fatty liver disease; RF, risk factor; T2DM, type 2 diabetes mellitus; WHR, waist-to-hip ratio; WC, waist circumference.