The Finnish Diabetes Risk Score (FINDRISC) and other non-invasive scores for screening of hepatic steatosis and associated cardiometabolic risk

Key messages

• Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in the general population and predisposes to cirrhosis, diabetes, and cardiovascular disease; screening apparently asymptomatic subjects for this condition is therefore warranted.

• A number of non-invasive scores combining clinical and biochemical variables have been developed and validated to predict hepatic steatosis.

• Compared to the other scores, the FINDRISC is equally accurate, simpler, and does not require any biochemical parameters, therefore being most suitable for office-based screening of hepatic steatosis.

• The FINDRISC has also been validated more consistently than other scores for predicting the risk of type 2 diabetes and cardiovascular risk, thereby allowing a more comprehensive evaluation of cardiometabolic risk of NAFLD subjects.

• Future research needs to validate scores to assess the risk of cirrhosis and liver-related complications of these subjects, which is currently best defined by liver histology.

Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in the world, affecting 20%–30% of the general adult population and 70% of obese and diabetic subjects (1). Growing evidence suggests that NAFLD patients experience a 1.5-fold higher overall mortality, deriving from liver-related and cardiovascular disease (CVD), and a 2-fold higher risk of developing diabetes over 10–15 years compared to the general population, independently of associated cardiometabolic risk factors (1). Although liver-related complications are largely confined to the progressive form of NAFLD, i.e. non-alcoholic steatohepatitis (NASH), a recent 5-year follow-up study showed that patients with simple steatosis can still develop NASH and fibrosis progression if they gain weight and have abdominal fat accumulation (2), suggesting that life-style-induced weight reduction may favourably impact not only cardiometabolic but also liver-related risk of these subjects (3). The individuation of apparently healthy subjects with hepatic steatosis would therefore represent an important public health problem and have preventive implications.

The definitive diagnosis of steatosis is radiological, with decreasing sensitivity of magnetic resonance spectroscopy (MRS), computed tomography (CT), and ultrasound (4); however, the most sensitive techniques (MRS and CT) cannot be adopted for population-based screening of fatty liver. Recently a number of non-invasive scores have been proposed to screen subjects at increased risk of steatosis for further radiological assessment (5–9) (Table I). These scores combine clinical and anthropometric variables, like age, gender, and indexes of adiposity, with biochemical parameters which are variably available in routine clinical practice, ranging from liver enzymes to insulin, apolipoprotein A-I, and α2-macroglobulin. The applicability of these scores for screening subjects with fatty liver varies widely across centres, largely depending on the local resources and protocols, and may be hardly suitable for a population-based screening. In this issue of Annals of Medicine, Carvalho et al. evaluate the Finnish Diabetes Risk Score (FINDRISC) as a screening tool for ultrasonographic steatosis (10). The FINDRISC was originally developed for the Finnish National Type 2 Diabetes Prevention Programme as a tool for primary health care workers to predict 10-year risk of developing type 2 diabetes without the need for laboratory tests (11). Accordingly, by using multivariate analyses, a score was assigned to seven clinical variables easily derived from history and clinical examination, yielding a sum ranging from 0 to 26 (Table I). In this study, a FINDRISC score ≥8 had a sensitivity, specificity, and positive and negative predictive values for ultrasonographic steatosis of 71%, 75%, 63%, and 81% and would mandate to further radiological testing 36% of a population with a steatosis prevalence of 37%.

Table I. Studies assessing non-invasive clinical and biochemical scores for detecting hepatic steatosis.

Author	Population	Non-invasive score	Components	Definition of steatosis	AUROC	Cut-off	Se	Sp
Bedogni (5)	496 Italian subjects, aged 18–75 y	FLI	BMI, waist circumference, triglycerides, GGT	u.s.	0.85	60	61	86
Poynard (6)	884 French patients, median age 49 y	Steatotest	ALT, α2-macroglobulin, apolipoprotein A-I, haptoglobin, total bilirubin, GGT, cholesterol, triglycerides, glucose, age, gender, BMI	Histology	0.80	0.50	0.72	0.75
Bedogni (7)	588 Italian subjects, aged 21–79 y	LAP	Waist circumference, triglycerides	u.s.	0.79	4.28^a	NA	NA
Lee (8)	10724 Korean subjects, mean age 52 y	HSI	Sex, BMI, AST, ALT, diabetes	u.s.	0.81	<30 or >36	0.93; 0.46	0.40; 0.92
Kotronen (9)	470 Finnish subjects (111 T2DM)	LFS	T2DM, metabolic syndrome, insulin, AST, ALT	MRS	0.87	−0.640	0.84	0.69
Carvalho (10)	821 Brazilian non-diabetic apparently healthy subjects, mean age 43 y	FINDRISC	Age, BMI, waist circumference, use of antihypertensive agents, history of hyperglycaemia, physical activity <4 h/wk, dietary intake of vegetables/fruits/berries	u.s.	0.80	8	0.71	0.75
		BMI		u.s.	0.83	26.6 kg/m^2	0.75	0.73
		Waist circumference		u.s.	0.86	93.6 cm	0.79	0.78
		FLI	see above	u.s.	0.86	NA	NA	NA
		LAP	see above	u.s.	0.83	NA	NA	NA

^aThe value of LAP is ln-transformed.

ALT = alanine aminotransferase; AST = aspartate aminotransferase; AUROC = area under the receiver operator curve; BMI = body mass index; FINDRISC = Finnish Diabetes Risk Score; FLI = fatty liver index; GGT = gamma-glutamyltransferase; HSI = hepatic steatosis index; LAP = lipid accumulation product; LFS = liver fat score; MRS = magnetic resonance spectroscopy; NA = not available; Se = sensitivity; Sp = specificity; T2DM = type 2 diabetes mellitus; u.s. = ultrasound.

Remarkably, the FINDRISC score performed better than liver enzymes alone, which could not predict steatosis (Table III of Carvalho et al. (10)), confirming that current cut-offs for liver enzyme elevation are insensitive and should be replaced by a strategy including overall evaluation of individual risk factors for NAFLD.

What are the limitations of this study? First, the predominant male sex of the study population and the relative insensitivity of the AUDIT tool for lower thresholds of alcohol intake let us suppose a proportion of male participants would have alcoholic, rather than non-alcoholic, steatosis. The under-estimated alcohol consumption of men could also account for the unexpectedly high (44%) prevalence of fatty liver in men, 4-fold higher than in women. Therefore, further validation in different, sex-balanced populations with more accurate detection of alcohol consumption is warranted. A second, obvious methodological shortcoming, acknowledged by the authors, relates to the insensitivity of ultrasound for mild-to-moderate (involving <33% hepatocytes) steatosis (4). However, I would remind any NAFLD expert that most prospective epidemiological studies connecting NAFLD to an increased cardiometabolic risk use a biochemical or ultrasonographic definition of fatty liver, and whether even minor degrees of liver fat infiltration substantially impact the risk of CVD and diabetes is currently unknown. In other words, the threshold of severity of steatosis we should pursue to impact individual cardiometabolic risk remains to be determined. The third limitation, intrinsic to the end-point of the study, is that assessing hepatic fat infiltration is not fully informative of the patient's risk of liver-related complications, including cirrhosis and end-stage liver disease, which is more closely related to the presence of NASH, with or without advanced fibrosis (1). For this aim, liver biopsy remains essential, but a number of non-invasive scores are being developed and will hopefully help select at-risk NAFLD patients for liver biopsy in the future (1).

On the other hand, the merits and clinical importance of this study far outweigh its limitations. First, this study validates a simple tool for prediction of hepatic steatosis. Compared to other scores, the FINDRISC is simpler, not requiring laboratory examinations, which makes it a practical tool for office-based screening of hepatic steatosis. Not less importantly, the FINDRISC may offer a comprehensive estimation not only of the risk of having fatty liver but also of the patient's overall cardiometabolic risk, as it has been more extensively validated than other scores in different populations for predicting the risk of developing diabetes and CVD (Table II) (12–25).

Table II. Studies assessing non-invasive scores of hepatic steatosis for predicting type 2 diabetes (T2DM) and cardiovascular disease (CVD).

Author (ref)	Population	Design (y)	Target condition	Cut-off	AUROC	Se	Sp
FINDRISC:
Lindstrom (11)	4746 Finnish subjects, aged 35–64 y	Prospective (10 y)	Drug-treated T2DM	9	0.85	0.78	0.77
Rathmann (12)	1353 German subjects, aged 55–74 y	Cross-sectional	T2DM by OGTT	9	0.65	0.82	0.43
Saaristo (13)	4622 Finnish subjects, aged 45–74 y	Cross-sectional	T2DM by OGTT	11	Men: 0.72; Women: 0.73	Men: 0.66; Women: 0.70	Men: 0.69 Women: 0.61
Franciosi (14)	1377 Italian subjects, aged 55–77 y	Cross-sectional	T2DM by OGTT	9	0.72	0.86	0.41
Silventoinen (15)	17725 Finnish subjects, aged 25–64 y	Prospective (9.9 y)	CHD, stroke, total mortality	4/5	Men: CHD 0.75, Stroke 0.73, Mortality 0.65; Women: CHD 0.81, Stroke 0.78, Mortality 0.74	NA	NA
Bergmann (16)	552 German high-risk subjects, aged 41–79 y	Prospective (3 y)	T2DM by OGTT	11	0.79	0.63	0.83
Lin (17)	2759 Taiwanese subjects, aged ≥18 y	Cross-sectional	T2DM (FPG ≥126 mg/dL)	9	0.73	0.67	0.67
Tankova (18)	2169 Bulgarian subjects with ≥1 risk factor for diabetes	Cross-sectional	Pre-DM and T2DM by OGTT	10	0.71	0.84	0.61
Makrilakis (19)	869 Greek individuals	Cross-sectional	T2DM by OGTT	15	0.72	0.82	0.60
Balkau (20)	3817 French subjects, aged 30–65 y	Prospective (9 y)	T2DM by FPG or antidiabetic treatment	9	Men: 0.68 Women: 0.81	Men: 0.70 Women: 0.80	Men: 0.55 Women: 0.66
Author (ref)	Population	Design (y)	Target condition	Cut-off	Results
FLI:
Gastaldelli (22)	1307 European non-diabetic subjects, aged 30–60 y	Cross-sectional	Carotid IMT	60	FLI correlated with carotid IMT independently of age, gender, recruitment centre, LDL, blood pressure, smoking
Balkau (21)	3811 European non-diabetic subjects, aged 30–65 y	Prospective (9 y)	T2DM	70	Multiple-adjusted OR for incident T2DM of 4.71 (95% CI 1.68–13.16) in men and of 22.77 (95% CI 6.78–76.44) in women, compared to a FLI <20
LFS:
Balkau (21)	3811 European non-diabetic subjects, aged 30–65 y	Prospective (9 y)	T2DM by FPG or antidiabetic treatment	−1.82	Multiple-adjusted OR for incident diabetes of 3.95 (95% CI 1.93–8.07)
LAP:
Kahn HS (23)	8767 US subjects, aged ≥18 y (NHANES III)	Cross-sectional	T2DM by FPG or antidiabetic treatment	66.1 (men); 60.4 (women)	After multiple adjustment, the upper quartile had OR 17 (95% CI 10–31) of T2DM compared to the lowest quartile
Bozorgmanesh (24)	3242 Iranian subjects, aged ≥20 y	Prospective (6 y)	T2DM by FPG, OGTT or T2DM by or antidiabetic treatment	NA	AUROC 0.71 (men); 0.78 (women 20–49 y), 0.65 (women ≥50 y)
Bozorgmanesh (25)	6751 Iranian subjects, aged ≥30 y	Prospective (8.6 y)	CVD	NA	After multiple adjustment, a 1-SD increment in lnLAP conferred an OR of 1.41 (95% CI 1.02–1.96) for incident CVD in women; no independent association in men

AUROC = area under the receiver operator curve; CHD = coronary heart disease; CVD = cardiovascular disease; FINDRISC = Finnish Diabetes Risk Score; FLI = fatty liver index; FPG = fasting plasma glucose; IMT = intima–media thickness; LAP = lipid accumulation product; LFS = liver fat score; NA = not available; OGTT = oral glucose tolerance test; OR = odds ratio; Se = sensitivity; Sp = specificity; T2DM = type 2 diabetes mellitus.

Importantly, the most widely used cut-off in these studies was 9, quite close to the 8 used in this study, further substantiating the emerging pathogenic links between hepatic fat infiltration and the development of diabetes and CVD (26).

Which is the optimal target of FINDRISC? When balancing costs and benefits, the optimal target population to apply this screening tool should probably be apparently healthy subjects, who might have steatosis and therefore be at risk for liver-related complications, cardiovascular disease, and diabetes. Subjects with an elevated FINDRISC score should go for further biochemical and radiological assessment and may be the target for prevention. Prevention should include primarily life-style intervention, which has been shown to reduce the risk of CVD in at-risk populations; furthermore, poor physical activity and weight gain have been associated with diabetes development and with progression from simple steatosis to NASH over 3–5 years in NAFLD subjects (2,27). Clearly, further large-scale population-based trials are warranted to assess the effectiveness and efficacy of such a preventive strategy against the development of liver-related and cardiometabolic morbidity and mortality. Future research needs also to validate non-invasive tools for predicting liver-related risk of these patients, to yield a thorough estimation of overall health-related risk of NAFLD patients (28).

Declaration of interest: The authors has no present or past conflict of interest to disclose.