Non‐alcoholic fatty liver disease and the risk of fibrosis in Italian primary care services

Abstract Background and aims The prevalence of non‐alcoholic fatty liver disease (NAFLD) is increasing globally. This study aimed to determine the prevalence of NAFLD and the probability of liver fibrosis in Italian primary care services. Methods We carried out a population‐based and nested case–control study including all individuals aged 18 years and above registered at Italian primary care services. Data were collected from the general practitioners' network from 2010 to 2017. NAFLD cases were identified via the ICD‐9‐CM and Hepatic Steatosis Index score > 36 and were matched each up to 10 controls. Other causes of liver diseases were excluded. The risk of fibrosis was assessed using the FIB‐4 and NAFLD fibrosis scores (NFS). Results NAFLD was present in 9% of the primary care population with high regional variability. Among NAFLD subjects: 25% had diabetes, 10% had chronic kidney disease, 11% had cardiovascular disease and 28% were obese. Furthermore, 30% had at least two comorbidities and 13% had cirrhosis. Once cirrhosis was excluded, the risk of any degree of fibrosis was 13.8% with NFS and 20.5% with FIB‐4 in subjects <65 years. Conclusions Even if there is an identification gap in primary care, recorded cases with NAFLD have a high frequency of associated comorbidities. Despite regional variability, a close relation between cirrhosis and NAFLD exists (OR: 3.48, 95% CI: 3.23–3.76). Therefore, the use of non‐invasive tests should be promoted in primary care as a useful tool for the early identification of fibrosis risk, independently of evidence of steatosis.


| BACKG ROU N D
Non-alcoholic fatty liver disease (NAFLD) is characterized by the excessive accumulation of triglycerides in the liver. It is often associated with insulin resistance and metabolic syndrome (MetS) with common clinical manifestations, such as hypertension, dyslipidemia, visceral adiposity and glucose intolerance. NAFLD is considered the most common cause of liver injury, with a prevalence exceeding 20% in the general population. 1 NAFLD's 'global burden' is linked to its potential evolution through inflammation into non-alcoholic steatohepatitis (NASH) and fibrosis.
Fibrosis is the determining factor in the natural history of NAFLD.
Fibrosis is significantly associated with mortality and morbidity from cardiovascular diseases, extra-hepatic cancer and major hepatological events (i.e., cirrhosis with its complications and hepatocellular carcinoma). 2 Furthermore, the early identification of patients with significant fibrosis within a population at risk and the correct management of precipitating factors may reduce the increasing request for liver transplantation owing to cirrhosis secondary to NASH. 3 Despite the great impact of fibrotic liver disease on the health systems, healthcare costs and quality of life, 4 no effective treatment has been approved for the management of NASH. The current prevalence of NAFLD and NASH in Italy is unknown. [5][6][7] Therefore, the early diagnosis of NASH is a highly recommended strategy. [8][9][10][11] Several NITs have been developed as an alternative to biopsy for diagnosing fibrotic NASH. In almost all the guidelines, only two NITs have been recommended for diagnosing NAFLD in subjects requiring liver consultation: the NAFLD fibrosis score (NFS) and the FIB-4 index. 12 Both can help general practitioners (GPs) to identify patients who need to be referred to a specialist for better stratification of liver disease.
Although NAFLD represents an emerging problem with a high impact on healthcare systems, it has been reported that many countries have not prepared specific plans or models of care for this chronic condition 13,14 and that there is a diagnostic/registration gap in primary care in Europe. 6 Our study aimed to determine the prevalence of NAFLD and the probability of fibrosis using NITs in the Italian adult primary care service. The secondary endpoints were the assessment of comorbidities and the predicting factors of NAFLD.

| MATERIAL S AND ME THODS
This is a retrospective study with de-identified cases. The Ethical Committee of Fondazione Policlinico Gemelli IRCCS approved the study (Study ID 51634, protocol no. 2545/19).

| Study population
All individuals aged 18 years or older with at least 1 year of clinical history in the database from 2008 to 2017 were included in the study.

| Case definition
Patients with NAFLD were identified through the ICD-9-CM diagnostic code (571.8) 17 and the Hepatic Steatosis Index (HSI) >36, 18 calculated using the last measured parameters. The date of diagnosis of NAFLD was considered the study index date.
The HSI score was obtained using the following formula:

Lay Summary
• Forecast epidemiological data suggest that the prevalence of NAFLD of 20-30% in the general population will cause a growing incidence of cirrhosis and liver cancer.
• The presence of NAFLD was observed in 9% of subjects in primary care; this finding underlines a registration gap likely owing to non-invasive diagnostic difficulties and patient identification.
• The presence of fibrosis is the main driver of hepatic and extra-hepatic complications. Moreover, the presence of fibrosis in NAFLD may remain undiagnosed, reducing treatment options because several physicians are unaware of non-invasive tests (NITs) based on a simple combination of laboratory results other than liver biopsy for stratifying the risk of liver fibrosis.
• This is the first study to evaluate the risk of fibrosis in Italian primary care. We provide evidence that strongly suggests identifying NAFLD subjects at risk for fibrosis.
• The findings suggest that using NITs in primary care services may optimize the appropriate referral of NAFLD subjects at risk for fibrosis.
A value of >36 was used to rule in steatosis and validated in the Italian population. 18,19 All cases with a history of alcohol abuse or alcohol-related diseases, hepatitis B or C infection, autoimmune liver diseases, primary biliary cholangiopathy or hepatic or extra-hepatic neoplasia before or on the index date were excluded. Details on the exclusion algorithm were reported in Table S1.

| Study population characteristics
The following data were retrieved for each patient: date of birth, age, gender, height, weight, blood pressure and laboratory data (transaminases, platelet counts, glycaemia, glomerular filtration rate [GFR]) assessed 2 years before or 6 months after the index date. The most recent results recorded in 2017 were considered.
Comorbidities, such as diabetes, MetS, cerebro-cardiovascular diseases, congestive heart failure and HIV, were identified using the ICD-9-CM codes (Table S2). Charlson Comorbidity Index has been used for comorbidities.
Cirrhosis was identified using the ICD-9-CM codes (before or after the index date), and/or the presence of oesophageal varices was identified using the ICD-9-CM codes (456.0 and 456.1).
Polypharmacy was defined as the regular use of at least five concurrent categories according to the WHO Anatomical Therapeutic Chemical Classification System, 20 prescribed within 3 months before or after the index date.

| Liver fibrosis assessment
The FIB-4 and NFS were used to assess the probability of the presence of liver fibrosis. 21,22,23 The scores were calculated as follows: Since the scores are influenced by age, subjects were divided into two groups: those under 65 years old and those 65 years or older. to identify the cases with a higher probability of advanced fibrosis. 12 Age-adjusted scores were used for subjects aged 65 years and older. A low probability of developing fibrosis was considered when the FIB-4 was <2 or NFS was <−0.12. 21

| Nested case-control analysis
The first cases of NAFLD identified in the eligibility period were matched to 10 controls for age (±5 years), sex, the database's entry date and the follow-up duration. This design allowed us to estimate the determinants of NAFLD.

| Data analyses
Descriptive statistics were carried out. Continuous and categorical variables were presented as means (standard deviation, SD), medians, interquartile ranges or absolute numbers and percentages, as appropriate. The prevalence of NAFLD in 2017 was calculated as the ratio of the number of cases divided by the total active HSD subjects in the same year, considering all cases, and subsequently stratified by gender, age, or region of residence. Trends of NAFLD incidence and prevalence during 2008-2017 were considered.
Differences in the demographic and clinical characteristics between the cases and controls were analysed using the t test. The chisquare was used to test for between-group differences among the continuous and categorical variables. A conditional logistic model was adopted to estimate the univariate and multivariate odds ratio (OR) and related 95% CI for each determinant of NAFLD. No imputation was provided for the missing values in the categorical variables and they were considered a separate category in each model.

| Characteristics of patients diagnosed with NAFLD
The prevalence of NAFLD among the active patients undergoing primary care (i.e., alive and registered in the GPs' lists) in HSD in 2017 is shown in Table 1. Among the 918 954 active individuals, 83 981 (9.14%) had NAFLD, 18309 were identified using the ICD-9-CM diagnostic code, and 65 672 had an HSI score of >36. (Table S3 and S4).
NAFLD was higher present in male than in female subjects (10.   Figure 2).

| Comparison of the clinical characteristics of the NAFLD cases and matched controls
Cases with NAFLD (n = 67 747) were matched to 252 991 controls ( Table 3) across all ages ( Figure 3A).
The BMI (as a continuous variable and obesity: BMI >30) was greater among the cases than in controls, while there was no difference in systolic and diastolic blood pressure. Except for HIV, the proportion of comorbidities was higher among the cases than in controls. The presence of cirrhosis and the complications of liver diseases were significantly higher among the cases than in controls  Figure 3D). In subjects aged >65 years, the probability of fibrosis was 31.6% in NAFLD cases and 42.6% in controls using the FIB-4 ( Figure 3E).

| Factors associated with the presence of NAFLD
When the conditional logistic regression was used ( We also performed a multivariable analysis matched by region (

| DISCUSS ION
The epidemiological models developed on NAFLD suggest that the incidence of liver-related complications will double in the next few years. 24 To our knowledge, this is the first study investigating the burden of NAFLD in the Italian primary care setting using real-world data. More than 94% of cases with NAFLD were identified using the HSI, which had been previously indicated as a predictor of NAFLD in the Italian population. 18,19 This observation shows that many cases go undetected and/or unreported, which could be ascribed to the lack of a specific ICD-9-CM code, thus placing GPs in a difficult position on how to record the condition and, consequently, how to best identify patients. 25 In recent years, scientific societies have focused on the awareness of NAFLD, especially in subjects with simple steatosis at high risk of progression to NASH because of the need for close monitoring and urgent therapeutic intervention. Owing to confusion over the term nonalcoholic, 11 From our study, the presence of NAFLD was in 9% of the active population in primary care in 2017, with significant regional variability that could be partially explained by social and cultural differences in diet and lifestyle. This variability was present in other studies on other chronic diseases. 28 Historical data on the prevalence of fatty liver in Italy from the Dionysos study are from a general population cohort in a different historical period, with different lifestyles and in the age group 12-65 years, and from an Italian regional district. 29 To the best of our knowledge, our study represents the first epidemiological study in Italian primary care services and also focuses on regional variability.
The low prevalence (9%) confirms the identification/registration gap at the primary care level. This phenomenon seems to be related not only to the low awareness of NAFLD; however, it may be ascribed to structural elements, such as the lack of the disease-specific code and the failure or partial recording in the electronic database of the data of patients who did not consent to share anonymized data for the calculation of the diagnosed indices above as well as the lack of recording of imaging parameters and liver function tests. The lack of data entry is certainly influenced by the low confidence in primary care in diagnosing NAFLD, which, in the absence of a specific code, is still not perceived as an important disease with the potential risk of evolution. Of course, it should be considered that guidelines on using NITs to identify cases at risk of fibrosis have only recently been published in Italy. 30 The progressive increase in the prevalence and incidence of NAFLD cases may be explained by the increased primary care physicians' focus on liver disease in programs to identify HCV cases for treatment with direct antiviral drugs. The reduction in incidence could be explained by the stable population in primary care. The database used in primary care is unable to identify individuals with NASH who are likely to be included in NAFLD diagnosis.
The diagnosis of NASH is based on the finding of steatosis, inflammation and ballooning on histological examination of liver biopsy.
The absence of an administrative code and specific therapy probably further reduces the possibility of identifying cases in primary care. This might be explained by the increased hospitalization rate or institutionalization of subjects who no longer receive GP consultations. A similar trend was observed in women, whereas in men, the prevalence increased up to 67 years. Until this age, the overall prevalence was higher in men and then became greater in women.
More than half of NAFLD patients had at least two coexisting chronic conditions. Diabetes (24.7%) and cerebrovascular and/or cardiovascular diseases (11%) were the most prevalent conditions identified respectively. These results are consistent with the findings from EU cohorts, and diabetes appears to have a significantly higher prevalence in NAFLD patients than in the general Italian population (8.3%). Our findings confirm recent observations on the UK biobank, which evaluated non-invasively that steatosis is associated with a higher risk of major cardiovascular events. 31 MetS is associated with NAFLD, confirming the importance of insulin resistance as the underlying causal factor in the absence of diabetes.
We obtained a correlation between severely impaired kidney function and NAFLD, as recently reported in a large meta-analysis 32 and a registry study where a reduction in the GFR was associated with an increase in FIB-4 and NFS values. 33 The high proportion of patients with polypharmacy was not surprising, as patients with NAFLD had concomitant comorbidities, TA B L E 3 Characteristics of the non-alcoholic fatty liver disease cases and related controls Description Cases (n = 67 747), n (%)/mean ± SD

TA B L E 3 (Continued)
especially those older than 65 years. 34 Since using several drugs may cause additional liver injury, special attention should be paid when prescribing a new drug to a patient with NAFLD.
We obtained a high prevalence of advanced compensated liver disease and/or compensated cirrhosis (13%) in cases with NAFLD.
This implies that an urgent, specific check and the promotion of actions devoted to the early identification of patients at risk of fibrosis using NITs is required, as scientific societies have already suggested. 27 The use of NITs in primary care may help GPs identify NAFLD subjects with significant fibrosis, especially those <65 years,   The increased proportion of fibrosis is responsible for both the progression of the disease and the increased mortality. 36 Approximately 40% of NAFLD subjects with fibrosis may progress to more severe forms, as observed in long-term follow-up studies. 37  TA B L E 5 Determinants of nonalcoholic fatty liver disease occurrence (matched by region) the fast identification of patients at risk for liver and non-liverrelated complications needing a specialistic consultation. Therefore, the role of GPs is crucial in correctly prescribing referrals to a liver specialist in this scenario.

FU N D I N G I N FO R M ATI O N
This work was supported by a project grant from Gilead Sciences

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable.