Introduction
In the UK, liver disease is third commonest cause of premature death.1 Nonalcoholic fatty liver disease (NAFLD) is present, often undiagnosed,1 in 30% of the UK population2 and is a risk factor for extrahepatic diseases such as type 2 diabetes, cardiovascular disease, chronic kidney disease,3,4 and increased long-term risk of developing cancer.5,6 Evidence shows that as fibrosis stage increases, liver-related mortality increases exponentially.7 We have shown recently that ∼20% of patients with a liver fibrosis stage of ≥F1 (≥6.0 kPa/low fibrosis) progressed to advanced fibrosis/cirrhosis during a 5 year period of follow-up.8 Therefore the detection of liver fibrosis is important because it is a key risk factor for cirrhosis, hepatocellular carcinoma and end stage liver failure.6,9
There are a growing number of liver fibrosis assessment services in primary care that use vibration-controlled transient elastography (VCTE) to identify patients who require specialist referral to hepatology services. In 2016, the National Institute of Health and Care Excellence (NICE) NAFLD Guidelines recommended the use of the enhanced liver fibrosis (ELF) test as part of a pathway for the identification of patients at high risk of advanced liver fibrosis.10 We developed this further11 and introduced a primary care liver pathway12 and Community Liver Service for GPs to refer patients with suspected severe liver fibrosis. There are uncertainties regarding the performance of ELF at predicting significant fibrosis (≥F2) in real-world practice and, although recommended by NICE, ELF is not widely available.
Other tests such as the NAFLD fibrosis score,13 FIB-414 and AST to platelet ratio index (APRI) score15 are less expensive within the NHS, but require measurement of aspartate aminotransaminase (AST), and AST is not routinely measured as part of the normal ‘liver function test’ panel. Thus, there is a need to offer an alternative method of evaluating patients at risk of liver disease without incurring the additional expense of ELF,16 or extra requirement and expense of measuring AST. The NICE guidelines recommended ELF cutoff value for predicting advanced fibrosis (≥F3) is 10.51.17 However, individuals with significant fibrosis (≥F2) are at substantially increased risk of type 2 diabetes, heart disease,18–21 cirrhosis and overall mortality.22,23 Detection of ≥F2 is difficult,24 and although there are several serum biomarkers available for the detection of liver fibrosis,25 no one biomarker test is recommended for the detection of ≥F2.
We conducted a retrospective evaluation to provide real-world findings for other healthcare providers contemplating implementing a similar service. This retrospective evaluation assesses how ELF test cutoffs perform in a real-world setting, and estimates the score with the optimum balance of sensitivity and specificity (Youden’s index)26 of ELF for identification of significant (≥F2) and advanced fibrosis (≥F3). We examined whether alanine transaminase (ALT), body mass index (BMI) and glycated hemoglobin (HbA1c), three widely available variables associated with liver disease, can be used as predictors of ≥F2.
Aims
To evaluate:
The optimum ELF cutoff value for predicting advanced (≥F3/≥9.7 kPa) fibrosis;
Whether ELF can predict significant (≥F2/≥8.2 kPa) fibrosis;
If routinely collected individual patient level data can predict ≥F2; and test whether they improve the performance of ELF to predict ≥F2; and
What factors: (a) are independently associated with ≥F2 liver fibrosis, and (b) predict liver fibrosis ≥F2.
Methods
We used a retrospective cohort of patients (derivation cohort) recruited from the Southampton Community Liver Service between Jan-Dec 2020. An independent cohort (validation cohort) of patients recruited to the liver service between Mar-Dec 2021 was used to validate an algorithm developed in the derivation cohort for identifying patients with liver fibrosis. Using the Southampton primary care liver pathway to identify at risk patients (Supplementary File 1), GPs referred patients to the Community Liver Service for VCTE assessment.
Inclusion criteria
Adults (≥18 years of age) with an ELF score of ≥9.0; an alcohol use disorders identification test (AUDIT)27 score of <14,27,28 (indicating low risk, hazardous and harmful alcohol consumption) and VCTE readings between 1.1 kPa-75.0 kPa.
Exclusion criteria
Individuals with incomplete data, patients entering the pathway with an ELF score <9.0, an alcohol use disorders identification test (AUDIT score of ≥15, indicating alcohol dependence),27,28 and those identified with chronic viral hepatitis, autoimmune liver disease, or haemochromatosis.
Data collection
VCTE assessment took place at a primary care site in Southampton. The FibroScan Mini+430 model with automated M and XL probe selection was used. Assessment took 20 minutes and was complete after 10 successive valid (IQR/MED <30%) measurements were obtained.
Data analysis
Excel, Excel Solver29 plug-in, SPSS statistics software (version 27), R version 3.4.4 (2018-03-15) were used. Data were cleaned and any incomplete data were excluded from this evaluation. 273/350 patients in the derivation cohort and 115/176 in the validation cohort were eligible for retrospective evaluation (Fig. 1).2728
Statistical analysis
Validated cutoff values were used for the ELF scoring system.17,30,31 Biopsy-confirmed thresholds, using the NASH CRN classification system, were used for the cutoff values for VCTE assessment for fibrosis (kPa) and steatosis (dB/m2, Supplementary Tables 1–3).32 Data were stratified by fibrosis stage, medication (statins/no statins), sex (male/female), diabetes status (diabetes/no diabetes), and BMI ≥30 kg/m2/BMI <30 kg/m2. Standard descriptive statistics were used to summarize variables: mean and standard deviation (±SD) for continuous variables or median and interquartile range (IQR) for non-normally distributed variables, and numbers and percentages for categorical variables. The chi-square test for independence (α=0.05) was used to determine the relationship between categorical variables. Two-tailed independent samples t-tests were used to compare the differences between groups and Fisher’s exact test was used, when n≤5, to determine if there was a significant association. The relationship between F2 and F0-F1 and F3-4 was evaluated using Kruskal-Wallis H test and Mann-Whitney U tests with Bonferonni adjustment. Backward elimination binary logistic regression analysis and receiver operator characteristic (ROC) curve analysis were used to (1) test the independence of associations between variables collected before VCTE assessment and liver fibrosis stage and (2) assess the risk prediction ability of variables to identify ≥F2 and ≥F3 as binary outcomes. The area under the receiver operator curve (AUROC) was used to compare the diagnostic accuracy of ALT, BMI, HbA1c, and ELF. The Obuchowski index was used to calculate a weighted AUROC to compare ELF to the biopsy-confirmed VCTE thresholds.32 The Obuchoswki index is explained in more detail in Supplementary File 2. Youden index analysis26 was applied to find the optimum cutoff value of ELF for ≥F2 and ≥F3. The DANA33 (difference between the mean fibrosis stage of significant (≥F2) fibrosis minus the mean fibrosis stage of nonsignificant (F0-F1) fibrosis) was applied according to the prevalence of fibrosis stages.
Table 1Characteristics of patients in the (a) derivation cohort and (b) validation cohort
| Patient characteristics | (a) Derivation cohort (n=273) | (b) Validation cohort (n=115) |
|---|
| Men sex, n (%) | 151 | 55.3 | 64 | 55.7 |
| Minority ethnic groups, n (%) | 65 | 23.8 | 19 | 16.5 |
| Median age, years (IQR) | 57 | 47–64 | 61 | 50–69 |
| Mean ELF score, (SD)F | 9.9 | 0.8 | 10.2 | 0.6 |
| Mean weight, kg (SD) | 90.2 | 20.2 | 93.7 | 19.9 |
| Median BMI, kg/m2 (IQR) | 30.8 | 27.7–35.2 | 31.6 | 27.4–36.4 |
| BMI≥30 kg/m2, n (%) | 167 | 61.2 | 70 | 60.9 |
| Diabetes positive, n (%)¶ | 110 | 40.3 | 31 | 26.9 |
| Mean HbA1c, mmol/mol, (SD) | 43.2 | 14.1 | 45.4 | 14.6 |
| ALT≥40 IU/L, n (%) | 153 | 56.0 | 58 | 50.4 |
| Mean ALT, IU/L (SD) | 52.47 | 37.4 | 44.1 | 24.0 |
| Mean VCTE reading, kPa (SD) | 9.0 | 7.8 | 8.6 | 6.2 |
| Mean CAP score, dB/m2 (SD) | 319.2 | 58.1 | 315.6 | 52.0 |
| High alcohol, n (%)B* | 65 | 24.0 | 26 | 22.6 |
| Smoker, n (%) | 45 | 16.5 | No data |
| Fibrosis stage | | | | |
| F0 (<6.0 kPa), n (%) | 113 | 41.4 | 47 | 40.9 |
| F1 (6.0–8.2 kPa), n (%) | 58 | 21.2 | 29 | 25.2 |
| F2 (8.2–9.6 kPa), n (%) | 25 | 9.2 | 10 | 8.7 |
| F3 (9.7–13.5 kPa), n (%) | 35 | 12.8 | 14 | 12.2 |
| F4 (≥13.6 kPa), n (%) | 42 | 15.4 | 15 | 13.0 |
| ≥F2, n (%) | 102 | 37.4 | 40 | 34.8 |
| ≥F3, n (%) | 77 | 28.2 | 31 | 26.9 |
| Steatosis grade | | | | |
| S0 (<302 dB/m2), n (%) | 90 | 33.0 | 42 | 37.2 |
| S1 (≥302 dB/m2), n (%) | 56 | 20.5 | 26 | 23.0 |
| S2 (≥331 dB/m2, n (%) | 15 | 5.5 | 4 | 3.5 |
| S3 (≥337 dB/m2, n (%) | 112 | 41.0 | 41 | 36.3 |
| Medication | | | | |
| Antidepressants, n (%) | 75 | 27.5 | 23 | 20 |
| Antihypertensives, n (%) | 116 | 42.5 | 53 | 46.1 |
| Anticoagulants, n (%) | 36 | 13.2 | 10 | 8.7 |
| GLP-1 agonist, n (%) | 13 | 4.8 | 2 | 1.7 |
| Statins, n (%) | 88 | 32.2 | 39 | 33.9 |
| AIIR blockers, n (%) | 22 | 8.1 | 7 | 6.1 |
Individual predictor variables
ALT,34 BMI35 and HbA1c36–38 are associated with liver fibrosis, AUROC was used to evaluate their combined performance in predicting significant (≥F2) and advanced (≥F3) fibrosis.
Algorithm
We combined BMI, HbA1c with ALT to develop an algorithm to predict the probability of a patient having ≥F2. A full description of the method is included in Supplementary File 3.
Validation data
Data from different patients referred to the Community Liver Service in 2021 were used to develop an independent validation cohort, to validate the algorithm developed from the derivation cohort. A description of the method is included in Supplementary File 4.
Results
Patient characteristics are presented in Table 1
Derivation cohort
Median (IQR) age was 57 years (47–64), 55.3% were men. Mean (±SD) VCTE reading and controlled attenuation parameter (CAP) scores were 9.0 kPa (±7.8) and 319.2 dB/m2 (±58.1), respectively. 24% (n=65) were consuming alcohol at harmful and hazardous levels,27,28 61.2% (n=167) had a BMI ≥30 kg/m2 and 40.3% (n=110) had diabetes.
Validation cohort
Median (IQR) age was 61 years (50–69), 55.7% were men. Mean (±SD) VCTE reading and CAP scores were 8.6 kPa (±6.2) and 315.6 dB/m2 (±52.0), respectively. Up to 22.6% (n=26) were consuming alcohol at harmful and hazardous levels,27,28 60.9% (n=70) had a BMI ≥30 kg/m2 and 26.9% (n=31) had diabetes.
Prevalence of liver fibrosis
Forty-two of the two hundred and seventy-three patients (15.4%) were identified as having advanced fibrosis/cirrhosis (F4/≥13.6 kPa), with 12.8% (n=35) having severe fibrosis (F3/9.7 to 13.5 kPa), 9.2% (n=25) having moderate fibrosis (F2/8.2 kPa to 9.6 kPa) and 62.6% (n=171) having no to low fibrosis (F0 to F1/<6.0 kPa/≥6.0 kPa to 8.1 kPa). The characteristics of patients by fibrosis stage are shown in Supplementary Table 4.
Factors associated with ≥F2 liver fibrosis
ELF, BMI ≥30 kg/m2, ALT ≥40 IU/L and HbA1c were all positively associated with significant (≥F2) fibrosis (p=0.001, p≤0.001, p=0.005 and p=0.002 respectively (Supplementary Table 5). The results for data stratified by sex, BMI, diabetes status, and medication are shown in Supplementary Tables 6–9 respectively.
Predictors of ≥F2
Median (IQR) BMI of patients with F0-F1 was 30.0 kg/m2 (26.0–32.8) and 32.0 kg/m2 (29.3–38.9) in patients with F2 (p=0.003). Mean (SD) HbA1c of patients with F0–F1 was 39.9 mmol/mol (12.0) and 48.5 mmol/mol (15.7) in patients with F2. In total, 26.3% (n=45) of F0-F1 patients and 64.0% (n=16) of F2 patients were diabetes positive (p<0.001) and 50.3% (n=86) of patients with F0-F1 and 76% (n=19) of patients with F2 had a BMI ≥30 kg/m2 (p=0.016) (Supplementary Tables 10a and b).
ELF
As a predictor of significant (≥F2/≥8.2 kPa) or advanced fibrosis (≥F3/≥9.7 kPa) ELF showed a fair performance, AUC=0.70, 95% confidence interval (CI: 0.64–0.76 and AUC=0.72, 95% CI: 0.65–0.79 respectively (Fig. 2). Applying the Obuchowski index showed a slight improvement in the estimated accuracy of ELF for identifying ≥F2 and ≥F3 (0.773 and 0.789 respectively), Supplementary Table 11. Youden’s index calculated ELF=9.85 for ≥F2 and ELF=9.95 for ≥F3. The 2020 and 2021 DANA scores (Supplementary Table 12) show that the prevalence of fibrosis is not evenly distributed across the five fibrosis stages, when compared to the uniform prevalence distribution DANA of 2.5. Missed cases are defined as patients whose VCTE reading showed they had significant fibrosis (≥F2) and their ELF score was <9.0 (2020 Community Liver Service threshold), <9.8 (manufacturers of ELF threshold for severe fibrosis)39 or <10.51 (threshold proposed by NICE).17Table 2 shows that when ELF<10.51 there are n=20 missed cases for F2, n=24 missed cases for F3 and n=25 missed cases for F4.32
Table 2Number of patients below the selected ELF score thresholds and their VCTE-confirmed fibrosis stage
| Fibrosis stage with VCTE thresholdsa | Total patients | ELF<9.0
| ELF<9.8
| ELF<10.51
|
|---|
| n | % | n | % | n | % |
|---|
| F2/≥8.2 kPa to 9.6 kPa | 25 | 1 | 4.0 | 8 | 32.0 | 20 | 80.0 |
| F3/≥9.7 kPa to 13.5 kPa | 35 | 1 | 2.9 | 9 | 25.7 | 24 | 68.6 |
| F4/≥13.6 kPa | 42 | 0 | - | 12 | 28.6 | 25 | 59.5 |
Individual variables
ALT alone showed a poor performance for predicting both ≥F2 and ≥F3, AUC=0.65, 95% CI: 0.59–0.72 and AUC=0.67, 95% CI: 0.61–0.74 respectively. BMI alone showed a fair performance for predicting both ≥F2 and ≥F3, AUC=0.72, 95% CI: 0.66–0.78 and AUC=0.71, 95% CI: 0.64–0.78 respectively. HbA1c alone showed a fair performance for ≥F2, AUC=0.70, 95% CI: 0.63–0.77 and a lesser performance for ≥F3 AUC=0.68, 95% CI: 0.610.76 (Supplementary Fig. 1).
Combining variables
As each of the individual variables (ALT, BMI, and HbA1c) did not show a good diagnostic performance for identifying liver fibrosis, we tested the effect of combining these variables. Diagnostic performance for identifying ≥F2 and ≥F3 improved when we combined ALT, BMI, and HbA1c, which had a good performance for identifying ≥F2 (AUC=0.80, 95% CI: 0.74–0.85) and a fair performance for identifying ≥F3 (AUC=0.78, 95% CI: 0.72–0.84, Fig. 3A). Adding ELF to the three variables increased the performance of ≥F3 to good (AUC=0.82, 95% CI: 0.76–0.88) and increased the performance of ≥F2 (AUC=0.82, 95% CI: 0.76–0.87, Fig. 3B). Although there was a trend toward an improvement in AUC with the addition of ELF, the differences in AUC were not statistically significant.
ALT, BMI, and HbA1c (ALBA) algorithm
The derivation cohort (n=273) was used to create the ALBA algorithm (Table 1). The equation for predicting ≥F2 was:
(ALT–28.826)*0.002638)+((BMI–23.291)*0.02152)+((HbA1c–28.462)*0.009975)
Applying the ALBA algorithm to the derivation data set also showed a good performance for predicting ≥F2 (AUC=0.80, 95% CI: 0.69–0.92, Fig. 4A).
Validation cohort
The validation cohort (Table 1), n=115, was used to validate the ALBA algorithm. Applying the ALBA algorithm to the validation cohort for predicting ≥F2 showed AUC=0.75, 95% CI: 0.66–0.85 (Fig. 4B).
ALBA and ELF
Diagnostic performance for identifying ≥F2 improved when we combined the ALBA algorithm and ELF. AUC=0.82, 95% CI: 0.77–0.88 for the derivation cohort and AUC=0.76, 95% CI: 0.67–0.86 for the validation cohort (Fig. 4C, D respectively).
Discussion
Summary
Our results show that when compared to validated VCTE cutoff values for the stages of liver fibrosis,32 the National Institute for Health and Care Excellence (NICE) recommended cutoff value (ELF≥10.51)17 for predicting advanced fibrosis (≥F3) is too high. Youden’s index shows the optimum cutoff value for ≥F3 in this population is an ELF=9.95, and for ≥F2 is an ELF=9.85. The NICE cutoff value therefore should be viewed as a recommendation as our study, and others,40,41 show that the ELF cutoff value should be set according to the population it is being used for. To evaluate the performance of ELF for identifying ≥F2 and ≥F3, we used the novel and underutilized Obuchowski index and the more standard AUC. We found the Obuchowski index shows a slightly higher performance than does AUC, although this increase does not change the performance classification of ELF. We have shown that referrals to the Community Liver Service have a high proportion of patients with obesity (BMI≥30 kg/m2) and type 2 diabetes, which led to the development of the ALBA algorithm, as an alternative method of evaluating patients at risk of liver disease. We validated the ALBA algorithm, compared the performance with ELF, and found that both offered a fair performance for predicting ≥F2. Importantly, combining ELF with ALBA improved the performance for predicting ≥F2. Our simple ALBA algorithm was not designed to replace existing validated markers of fibrosis, but it could be a tool for GPs, who do not have access to these costly tests, to use to assess whether a patient is at risk of ≥F2.
Strengths and limitations
This study has shown that routinely available data can assess a patient for ≥F2. This study has also provided data to show liver disease is highly prevalent among patients with diabetes and/or BMI≥30 kg/m2.42–44
There were limitations to this study. This evaluation did not differentiate between NAFLD and alcohol related liver disease. Our sample size was small and there may have been some slight overfitting. Our data was not evenly distributed across the five fibrosis stages but represented a more realistic prevalence of fibrosis in a community setting. We did not have measurements of AST available, so we could not calculate other liver fibrosis scores such as the Fibrosis-414 score for comparison with ELF or ALBA. Finally, VCTE assessment is a validated noninvasive test used to measure liver stiffness,32 and although liver biopsy continues to remain the gold standard in the assessment for liver disease,45 it is invasive, costly and prone to sampling error.46 Moreover, liver biopsy is not feasible within a large community-based liver service that does not have the capability to monitor patients for any length of time post liver-biopsy procedure.
Comparison with existing literature
Previous studies have focused on patients with established NAFLD or screening for patients with advanced fibrosis/cirrhosis.47,48 However, it is early detection of NAFLD and early stage liver fibrosis (F2), an established risk factor for cirrhosis and overall mortality,49,50 that is key to helping prevent, control, and manage disease progression.
Our findings revealed that 40.3% of patients referred to the Community Liver Service had diabetes, six times higher than the prevalence of diabetes in the UK.51 Diabetes is important risk factor for NAFLD,37 yet liver function tests are not recommended in the NICE guidelines for diabetes.52 NAFLD is one of the most common causes of hepatocellular carcinoma and is likely to continue as the incidence of both obesity and type 2 diabetes continue to increase.53
Implications for practice
Health care providers considering implementing a liver service should consider a suitable ELF threshold to achieve the desired performance.41 This evaluation provided the Southampton Clinical Commissioning Group with the evidence needed to refine the primary care liver pathway ELF cutoff value, referral for VCTE assessment is now set to ELF≥9.5.
Up to 12.8% (n=25) of patients discharged back to their GP were found to have F2, a stage of liver fibrosis which puts them at an increased risk of type 2 diabetes and heart disease.18–21 Because we do not know what specific factors will predict disease progression, these patients need to be managed by their GP on the assumption that their liver fibrosis will progress.8
Conclusion
This study has shown that in the absence of access to noninvasive blood tests, the ALBA algorithm can predict the probability of a patient having ≥F2, a stage of fibrosis that can be treated with low doses of prescribed GLP-1 receptor agonists.22,23 We have further shown that combining ALBA and ELF improves risk prediction for ≥F2. Finally, this study highlights the disproportionate number of patients with diabetes and/or a BMI≥30 kg/m2 with liver fibrosis, which lends further weight to targeting these known high risk groups in screening for liver disease.
Supporting information
Supplementary File 1
Primary care liver pathway (abridged).
(DOCX)
Supplementary File 2
Obuchowski index.
(DOCX)
Supplementary File 3
Development of algorithm.
(DOCX)
Supplementary File 4
Validation cohort.
(DOCX)
Supplementary Table 1
ELFa test thresholds and predicted severity of liver fibrosis.
(DOCX)
Supplementary Table 2
VCTE cutoff values and liver stage fibrosis.32
(DOCX)
Supplementary Table 3
CAP cutoff values and grade of steatosis.32
(DOCX)
Supplementary Table 4
Characteristics of patients by fibrosis stage.
(DOCX)
Supplementary Table 5
Patient characteristics and their relationship with significant (≥F2) liver fibrosis.
(DOCX)
Supplementary Table 6
Characteristics of patients stratified by sex.
(DOCX)
Supplementary Table 7
Characteristics of patients stratified by BMI < and ≥30 kg/m2.
(DOCX)
Supplementary Table 8
Characteristics of patients stratified by diabetes status.
(DOCX)
Supplementary Table 9
Characteristics of patients stratified by statin prescribing.
(DOCX)
Supplementary Table 10
(a) A comparison of the characteristics of patients with no-low fibrosis (F0-F1) vs. patients with moderate fibrosis (F2). (b) A comparison of the characteristics of patients with advanced fibrosis (≥F3) vs. patients with moderate fibrosis (F2).
(DOCX)
Supplementary Table 11
Evaluation of the diagnostic performance of ELF using area under the curve (AUC) and the Obuchowski index (full calculations shown in Supplementary File 5a).
(DOCX)
Supplementary Table 12
Comparison of the difference in significant (≥F2) and nonsignificant (F0-F1) fibrosis means (DANA) in the 2020 and 2021 datasets (full calculations shown in Supplementary File 5b).
(DOCX)
Supplementary Fig. 1
Area under the curve (AUC) receiver operating characteristics (ROC) for the prediction of significant fibrosis (≥F2/≥8.2 kPa) and advanced fibrosis (≥F3/≥9.7 kPa) using (A) ALT; (B) BMI and (C) HbA1c.
(DOCX)
Abbreviations
- ALBA:
Alanine transaminase, body mass index and alanine transaminase
- ALT:
Alanine transaminase
- APRI:
Aspartate transaminase to platelet ratio index
- AST:
Aspartate transaminase
- AUC:
Area under the curve
- AUDIT:
Alcohol use disorders identification test
- BMI:
Body mass index
- CAP:
Controlled attenuation parameter
- CI:
Confidence interval
- CVD:
Cardio vascular disease
- DANA:
The difference between the mean fibrosis stage of advanced fibrosis minus the mean fibrosis stage of non-advanced fibrosis
- ELF:
Enhanced liver fibrosis test
- FIB-4:
Fibrosis-4 index
- GLP-1:
Glucagon-like peptide-1
- GP:
General practitioner
- HbA1c:
Glycated hemoglobin
- IQR:
Interquartile range
- M:
Medium
- METAVIR:
Meta-analysis of histological data in viral hepatitis
- NAFLD:
Nonalcoholic fatty liver disease
- NASH:
Nonalcoholic steohepatitis
- NASH CRN:
Nonalcoholic steahepatitis Clinical Research Network
- NFS:
NAFLD fibrosis score
- NHS:
National Health Service
- NICE:
National Institute for Health and Care Excellence
- NPV:
Negative predictive value
- PPV:
Positive predictive value
- ROC:
Receiver operating characteristic
- SD:
Standard deviation
- T2DM:
Type 2 diabetes
- VCTE:
Vibration-controlled transient elastography
- XL:
Extra large
Declarations
Acknowledgement
Southampton Clinical Commissioning Group and Solent Medical Services Southampton. For the purpose of Open Access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission. The authors would like to thank the NIHR Southampton Biomedical Research Centre and the University of Southampton for their support.
Ethical statement
This retrospective evaluation of the Southampton Community Liver Service used routinely collected data. All the data collection and analysis was conducted by the clinical team involved in delivering patient care. This evaluation was approved by the clinical lead for hepatology services at University Hospital Southampton and was registered for clinical audit (registration number: ZAUD7162) but not subject to review by an independent ethics committee and individual patient consent was not sought. All activities were performed following the guidelines of the Helsinki Declaration.
Data sharing statement
No additional data are available.
Funding
CDB and RMB are supported in part by the Southampton NIHR Biomedical Research Center (NIHR203319), UK.
Conflict of interest
The authors have no conflict of interests related to this publication.
Authors’ contributions
Study concept and design (TR, MM, JP) acquisition of data (TR), analysis and interpretation of data (TR, DF, CB), drafting of the manuscript (TR, CB), critical revision of the manuscript for important intellectual content (TR, DF, JP, MM, RB, CB) study supervision (CB). All authors have contributed to this study and have approved the final manuscript.