Obesity and type 2 diabetes are important risk factors underlying previously undiagnosed cirrhosis in general practice: a cross‐sectional study using transient elastography

Rising cirrhosis incidence and mortality in the United Kingdom has been attributed predominantly to excess alcohol consumption. However, metabolic risk factors such as Type 2 diabetes and obesity may also be important.


| INTRODUCTION
Chronic liver disease continues to be a significant public health burden. In England, diagnosed cirrhosis incidence has increased by 50.6% between 1998 and 2009, 1 and subsequently cirrhosis is now the third commonest cause of premature mortality in persons aged 20-54. 2 Population level alcohol consumption is still considered the major driver of liver-related mortality, and up to 62% of detected cirrhosis cases in the population are attributed to alcoholic liver disease (ALD). 1,3 However, the rising population prevalence of metabolic syndrome risk factors means that non-alcoholic fatty liver disease (NAFLD) and related cirrhosis are becoming a more significant issue. Obesity prevalence in the United Kingdom has risen by 54% in men and 65% in women between 1993 and 2012, and the prevalence of Type 2 diabetes is forecast to double between 2000 and 2030. Non-alcoholic fatty liver disease is therefore likely to become the leading cause of liver cirrhosis in the near future, and indeed in the US NAFLD has already overtaken alcoholic liver disease as a cause of listing for liver transplantation. 4 Several studies have demonstrated the important synergism between body mass index and alcohol consumption, and the increased risk of progressive fibrosis, 5 cirrhosis diagnoses 6 and liver-related death. 7,8 However, as NAFLD is discounted as a diagnosis in the presence of a specific cut-off of alcohol consumption (such as a weekly consumption of >14 units in women and >21 units in men), all cirrhosis in those with an alcohol consumption above the cut-off will be coded as alcoholrelated cirrhosis. 9 The overall impact of obesity, Type 2 diabetes and other metabolic risk factors on cirrhosis incidence and liver-related outcomes is therefore likely to be greatly underestimated by current practice.
We have previously demonstrated that a community-based service using transient elastography (TE) can detect appreciable quantities of previously unrecognised liver disease among people with hazardous alcohol use and people with Type 2 diabetes. 10 We extended this study with additional recruitment, to study the association of alcohol and metabolic risk factors with chronic liver disease in a community setting. The aims of this study are to characterise new clinically significant liver disease and cirrhosis within the screened population, and to identify the risk factors associated with elevated liver stiffness and cirrhosis.

| Study setting
This was a cross-sectional study with recruitment from four general medical practices in Nottingham, United Kingdom. Of these, two primary care medical practices were located in an affluent suburban borough, whilst the remaining two practices were situated in pre- clinician selects codes to record directly into the computerised medical record system. Local regulatory approval was obtained from the Leicester Research Ethics Committee (study identification 13/EM/ 0123), and written informed consent was gained from patients.

| Patient selection
Patient selection for this study has been previously published. 10 Briefly, adult patients (defined as aged 18 years or older) with selected risk factors for lifestyle related chronic liver disease were identified directly from the general medical practice electronic patient records. The electronic record search was performed at the commencement of the study. The studied risk factors were: In addition, 3% of the study patients from the 4 studied practices were patients with persistently elevated serum alanine aminotransferase (ALT) levels (>35 IU/L for women, >45 IU/L for men), who had neither hazardous alcohol intake nor Type 2 diabetes, and negative liver autoimmune and serological tests.
Patients were excluded from the study if (1) there was definitive evidence of hepatic fibrosis or cirrhosis already identified from previous investigations, (2) there was a contraindication to transient elastography (pregnancy, indwelling cardiac device), (3) they had metastatic malignancy, (4) they were unable to consent to investigation due to significant cognitive impairment or (5) they were housebound and could not attend the community practice. In addition, patients who presented with symptoms of decompensated liver cirrhosis (eg, jaundice, variceal bleeding, ascites) during the study period were excluded and instead triaged straight to urgent hospitalbased care rather than being screened using TE in primary care.

| Liver stiffness measurement
The methodology for invitation and screening of patients from the suburban medical practices has been published previously, and HARMAN ET AL.
| 505 included an initial screening blood biomarker prior to transient elastography. 10 Patients from the inner city practices were invited directly for an appointment to undergo transient elastography. The scan was performed by one of three nurses, all of whom had received formal training in liver stiffness measurement and had previously performed more than two hundred liver stiffness acquisitions in the hospital setting. Our nurse led transient elastography service has been established since 2009 and we have published the evaluation of this service in 2012 showing only 5.3% of scans were unreliable. 12 Patients were advised to be fasted for the examination. 13 Patients with a body mass index (BMI) of <35 kg/m 2 underwent TE examination with the Fibroscan FS402 device (Echosens) M probe in the general practice setting. Due to a high risk of unreliable or failed liver stiffness acquisition with body mass index (BMI) measures above this threshold demonstrated in previous studies, 14 patients with a BMI>35 kg/m 2 , and those with an initial failed liver stiffness acquisition using the M probe, underwent transient elastography in the hospital using the Fibroscan FS502 device XL probe.
Liver stiffness acquisition failure was defined as inability to obtain 10 valid liver stiffness measurements. Participants with failure of liver stiffness acquisition were excluded from the analysis. A successful acquisition was deemed unreliable if liver stiffness was ≥7.1 kPa and the interquartile range/median ratio was greater than 0.3 as per manufacturer guidance. 15 A liver stiffness threshold of 8.0 kPa or greater was used to define elevated liver stiffness, and hence clinically significant liver disease, in keeping with a previous large general population study in France 16 in which this cut-off was shown to be an accurate predictor of liver fibrosis on biopsy. The same liver stiffness threshold was used for patients with both alcohol and non-alcohol related liver risk factor as although alcohol-related liver fibrosis potentially results in slightly higher liver stiffness results than NAFLD, a recent metaanalysis of previous hospital studies has shown good accuracy for predicting significant (F2) liver fibrosis in patients with alcoholic liver disease using very similar liver stiffness cutoffs. 17 Patients with elevated liver stiffness results, including high but unreliable acquisitions, were reviewed by a visiting consultant hepatologist in the community (one of authors SR, EW, MJ, GPA or ING).
Alternative causes of chronic liver disease (eg, viral and autoimmune liver disease) were tested for. In addition, at the clinical discretion of the reviewing hepatologist, further investigations including ultrasonography, liver biopsy and endoscopy were arranged on a caseby-case basis.
Following the transient elastography appointment, patient's electronic primary care records were retrospectively examined to collect recent relevant clinical, anthropometric and laboratory test data (definitions of these risk factors are shown in Online Appendix S1). As waist circumference was not routinely measured, obesity was defined as the presence of body mass index ≥ 30 kg/m 2 . Subsequently, metabolic syndrome was defined according to the International Diabetes Federation definition as presence of obesity with 2 or more metabolic risk factors (hypertension, impaired fasting glucose or Type 2 diabetes, raised triglycerides or lowered high density lipoprotein cholesterol). 18

| Cirrhosis detection and associated risk factors
Cirrhosis was diagnosed clinically at the discretion of the visiting consultant hepatologist. To increase the positive predictive value of cirrhosis diagnoses, given the need for future cirrhosis surveillance investigations, cirrhosis diagnoses were not based upon an elevated liver stiffness measurement alone. Rather cirrhosis diagnoses were assigned using elevated liver stiffness measures in combination with either histological evidence of cirrhosis, endoscopic evidence of portal hypertension or ultrasound evidence of cirrhosis or portal hypertension (ie, nodular liver surface, splenomegaly or reversal of portal vein flow). Cirrhosis diagnoses were classified as alcoholic liver disease if hazardous alcohol use was present in the absence of obesity or Type 2 diabetes, as non-alcoholic fatty liver disease in the presence of Type 2 diabetes or obesity, but without hazardous alcohol use, and as dual aetiology if a combination of hazardous alcohol use and Type 2 diabetes or obesity was present. The number and aetiology of cirrhosis diagnoses in the general practice population before study commencement were obtained by searching the electronic patient records.
To evaluate the impact of the defined clinical and metabolic risk factors on the presence of significant liver disease, we compared the percentage of elevated liver stiffness and cirrhosis cases for patients with and without these risk factors. Subgroup analyses examining patients with hazardous alcohol use, Type 2 diabetes and both hazardous alcohol and Type 2 diabetes were performed.

| Statistics
Statistical analysis was performed using STATA version 13.1 (STATACORP LP, USA). Categorical data are presented as number (percentage).
Continuous data are presented as medians (range) (as all were nonnormally distributed). Demographic, anthropometric and laboratory test data were compared between patients with and without cirrhosis using the Mann-Whitney test as appropriate. Categorical variables were compared using chi-squared test or Fisher's exact test where appropriate.
To further evaluate the association of clinical and metabolic risk factors with clinically significant liver disease, for those risk factors which were associated with both presence of elevated liver stiffness and cirrhosis we report univariate odds ratios and 95% confidence intervals comparing patients with and without these clinical features in each of our studied groups.

| Study population
The total adult population in the studied primary care centres at commencement of the study was 20 868 patients (Table 1) Compared to non-attenders, transient elastography appointment attenders were significantly less likely to be hazardous alcohol users (43.6% vs 83.0%; P < .001), significantly less likely to be male (65.7% vs 71.7%; P = .004) and were significantly older (mean age 59.1 years vs 47.8 years; P < .001).  (Table 2). However, both hazardous alcohol use prevalence (47.3% vs 32.6%; P < .001) and median alcohol consumption (8 vs 3 units of ethanol per week; P < .001) were significantly lower in patients with elevated liver stiffness.

| Elevated liver stiffness and cirrhosis diagnoses
Prior to study commencement, there were 23 diagnosed cases of liver cirrhosis in the population of the studied general practices who were excluded from study. All cases had been diagnosed on the basis of histological evidence of liver cirrhosis, or presentation to hospital with decompensated liver cirrhosis. Their cirrhosis aetiologies were alcoholic liver disease (14 patients), Hepatitis B or C (5),

NAFLD (1) and other (3).
During the study, 209 patients with elevated liver stiffness attended and were reviewed in hepatology clinics and 27 of these were newly diagnosed with liver cirrhosis during the study period (3% of valid liver stiffness results). This, therefore, more than doubled the number of cirrhosis diagnoses in the studied general prac-   Normally distributed numerical variables are displayed as mean (standard deviation [SD]) and compared using the t test; non-normally distributed numerical variables are displayed as median (interquartile range) and compared using the Mann-Whitney test. Categorical variables are displayed as n (%) and compared using Fisher's exact test. P values ≤.05 are displayed in bold. ALT, alanine aminotransferase, kPa, kilopascals, LSM, liver stiffness measurement. a P value comparing normal and elevated liver stiffness in patients with hazardous alcohol use. b P value comparing normal and elevated liver stiffness in patients with Type 2 diabetes. c P value comparing normal and elevated liver stiffness in patients with both of these risk factors.

| Strengths and limitations
T A B L E 5 Odds ratios for presence of (a) elevated liver stiffness and (b) cirrhosis comparing presence and absence of obesity, metabolic syndrome and raised ALT level in patients with hazardous alcohol use (n = 391), Type 2 diabetes (n = 543) or both hazardous alcohol and Type 2 diabetes (n = 64) Hazardous alcohol use (n = 391) Type 2 diabetes (n = 543) Hazardous alcohol and Type 2 diabetes (n = 64) (a) Variable Exposed Non-exposed OR (95% CI) Exposed Non-exposed OR (95% CI) Exposed Non-exposed OR (95% CI) Hazardous alcohol and Type 2 diabetes (n = 64) (b) Variable Exposed Non-exposed OR (95% CI) Exposed Non-exposed OR (95% CI) Exposed Non-exposed OR (95% CI) ALT, alanine aminotransferase; CI, confidence interval, OR, odds ratio; odds ratios reaching statistical significance are displayed in bold.
believe is the clinical confirmation of cirrhosis diagnoses, for which we used additional supportive radiological, histological and endoscopic evidence. Although recent guidance has suggested that a liver stiffness reading of greater than 15 kPa is strongly suggestive of compensated cirrhosis, 19 the positive predictive value of liver stiffness readings in the community may be lower than previous secondary care studies due to reduced disease prevalence. We therefore believe the cirrhosis diagnoses we have reported are robust.
One of the limitations however is that only 45% of the eligible population underwent the transient elastography examination, although this is a greater response rate than the other major UK primary care liver stratification study thus far reported, which enrolled only 35% of patients defined as at risk from their alcohol consumption. 20 There was however a response bias with screening attenders being older, more female and with a differing proportion of hazardous alcohol use and Type 2 diabetes than non-attenders. It is therefore difficult to predict the incremental increase in cirrhosis which would be diagnosed if everyone invited had attended. A further limitation is that as we targeted only Type 2 diabetes and alcohol misuse as risk factors, though we have been able to show that obesity is an important co-factor in each, we are unable to assess whether it is an important risk factor in its own right. Based upon this, it is likely that we have screened the highest risk patients with obesity within the population, but we will have not detected patients with clinically significant liver disease and obesity alone as a risk factor.

| Comparison with other studies
Previous systematic reviews have highlighted the accuracy of transient elastography for stratifying fibrosis stage in secondary care populations with non-alcoholic fatty liver disease. 21

| Implications
In real life clinical practice, patients will have multiple risk factors for chronic liver disease. For example, within this study 44% of our population undergoing TE had two or more of hazardous alcohol use, Type 2 diabetes or obesity. Given this, and the synergism of these risk factors that the study highlights we feel that strictly dichotomising patients with alcoholic liver disease and NAFLD on the basis of a specific cut-off of alcohol consumption may serve to mislead. In our study patients with hazardous alcohol use, Type 2 diabetes and obesity had a cirrhosis prevalence of 13.3%, and elevated liver stiffness was seen in 46.7%. However, when considering Type 2 diabetes as the only selection criteria the equivalent figures were 3.7% and 31.5%. That the vast majority of significant liver disease detected was found in those with diabetes and/or obesity shows we think that alcohol alone is not likely to remain the cause of the majority of the emerging epidemic of liver disease in the United Kingdom in future.
Given the high prevalence of both elevated liver stiffness and cirrhosis detected in patients with Type 2 diabetes, in addition to proof of concept of screening such patients using transient elastography in primary care, we believe that formal screening for liver disease in Type 2 diabetes should now be considered. In addition to lifestyle alterations, early detection will allow optimal medication management of these individuals with antidiabetic medication which will also treat non-alcoholic steatohepatitis, such as pioglitazone 32 or liraglutide. 33,34 One option to increase the cost efficacy of this is for selective screening and surveillance based upon the type and number of risk factors, to ensure a high pre-test probability of identifying clinically significant liver disease. In this study we have demonstrated that both hazardous alcohol use and obesity are potentially useful markers of higher risk groups for this purpose. Further optimisation of community-based liver disease detection strategy is required, both in terms of initial risk factor selection, and whether the addition of other simple liver fibrosis tests to primary care algorithms improves the classification of liver fibrosis. 35 Such work though we feel is now a priority to enable the eventual roll out of selective screening to detect liver disease at an earlier stage.

| SUMMARY
Using Transient Elastography to selectively screen for liver disease in primary care, and referring in high risk patients to a hepatology clinic resulted in a more than doubling of the number of cirrhosis cases diagnoses in the studied population. The majority of newly identified cirrhosis cases had Type 2 diabetes and obesity as risk factors (and therefore, presumably had non-alcoholic fatty liver disease). The risk of cirrhosis was far higher in those with multiple risk factors (hazardous alcohol use, obesity and Type 2 diabetes). Focussing upon the combination of liver disease risk factors is likely to be the most effective way of designing cost effective investigation algorithms, and relevant interventions, for patients in primary care.

ACKNOWLEDG EMENTS
Declaration of personal interests: All authors declare no support from any organisation for the submitted work, no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years, no other relationships or activities that could appear to have influenced the submitted work.
Declaration of funding interests: Funding for study was provided Sciences Network (AHSN). The study sponsor is the University of Nottingham, who is data custodians but had no role in the design, analysis or interpretations of the data. All authors declare that they are free from other sources of external funding related to this study.