Article Text

Original research
Association of fat-to-muscle ratio with non-alcoholic fatty liver disease: a single-centre retrospective study
  1. Fengqin Yan,
  2. Guqiao Nie,
  3. Nianli Zhou,
  4. Meng Zhang,
  5. Wen Peng
  1. Department of General Practice, Union Hospital, Tongji Medical College of Huazhong University of Science and Technology, Wuhan, China
  1. Correspondence to Professor Wen Peng; pengwen666{at}sina.com

Abstract

Objectives Sarcopenia is a known risk factor for non-alcoholic fatty liver disease (NAFLD). Studies evaluating the association between the fat-to-muscle ratio (FMR) and NAFLD are limited. Therefore, the aim of our study was to investigate the association between FMR and NAFLD.

Design A retrospective study was conducted on individuals who underwent health examination at Wuhan Union Hospital between January 2020 and November 2021. Clinical data were collected from electronic medical records.

Setting Wuhan Union Hospital, Wuhan, China.

Participants 1592 participants aged ≥40 years who underwent body composition analysis and liver ultrasonography were retrospectively reviewed.

Outcome measures Liver ultrasonography was used to assess liver steatosis, and the fibrosis-4 index was used to calculate the risk scores for liver fibrosis. The 10-year atherosclerotic cardiovascular disease (ASCVD) risk prediction model was used to calculate ASCVD risk scores.

Results The FMR was significantly higher in individuals with NAFLD than in those without NAFLD (p<0.001). The prevalence of NAFLD gradually increased from FMR tertile 1 (reference) to tertile 2 (OR=1.49, 95% CI 1.13 to 1.97) and tertile 3 (OR=2.85, 95% CI 2.08 to 3.90). In addition, patients with NAFLD in FMR tertile 3 had a significantly higher risk of liver fibrosis (OR=4.48, 95% CI 2.12 to 9.50) and ASCVD (OR=4.63, 95% CI 2.62 to 8.19) than those in FMR tertile 1 after adjustment for multiple confounders.

Conclusion In this study, we found a significant association between FMR and NAFLD. A higher FMR indicates a higher risk of NAFLD in the study population and a higher risk of liver fibrosis and ASCVD in NAFLD patients.

  • Risk management
  • Obesity
  • PUBLIC HEALTH
  • HAEMATOLOGY

Data availability statement

Data are available upon reasonable request. Extra data can be accessed via the Dryad data repository at http://datadryad.org/ with the doi:10.5061/dryad.7d7wm3809.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THIS STUDY

  • Generalised estimating models were used to analyse the risk of liver fibrosis and atherosclerotic cardiovascular disease.

  • We adjusted for many confounding factors such as sex, age, tobacco use, alcohol use, hypertension, diabetes, transaminases, lipids and uric acid.

  • Data on insulin resistance, adipokines and myokines were not available in the study, so the underlying mechanisms could not be elucidated.

  • This is a retrospective cross-sectional study and cannot establish a causal relationship between fat-to-muscle ratio and non-alcoholic fatty liver disease.

Introduction

Non-alcoholic fatty liver disease (NAFLD) has become a public health problem affecting nearly a quarter of the world’s adult population.1 Due to excess calories and a sedentary lifestyle, the prevalence of NAFLD in China increased dramatically from 18% to 29% between 2008 and 2018, more than two times the rate of increase in Western countries.2 In a recent systematic review and meta-analysis, Riazi et al collected 72 studies with publication years from 2002 to 2021 to analyse the global prevalence of NAFLD and found that the prevalence of NAFLD increased significantly over time, rising steadily from 25.5% in 2005 or earlier to 37.8% in 2016 or later, and the prevalence of NAFLD in China was 32.5% from 19 studies published from mainland China.3 NAFLD encompasses a spectrum of liver disease ranging from steatosis to steatohepatitis, which can progress to cirrhosis and hepatocellular carcinoma. It is also associated with metabolic abnormalities, cardiovascular disease (CVD), extrahepatic tumours and an increased risk of all-cause mortality.4 Despite the economic burden of NAFLD on the healthcare system, there is a lack of definitive risk assessment and treatment for NAFLD. Further research is needed to improve the identification of high-risk individuals and to develop effective management strategies.

Risk factors for NAFLD include obesity, type 2 diabetes, genetic variants, obstructive sleep apnoea, intestinal dysbiosis and sarcopenia.5 Obesity, defined as abnormal or excessive fat accumulation, is a well-established risk factor for NAFLD.6 Sarcopenia is also a newly recognised risk factor in the development and progression of NAFLD.7 8 A previous study suggested that abnormal body composition, such as increased fat mass accompanied by reduced muscle mass, represents a dual metabolic burden that may lead to an increased risk of insulin resistance (IR).9 IR has been implicated in the pathogenesis of NAFLD along with oxidative stress, dietary factors and genetic factors.10 Several studies have reported that the ratio of skeletal muscle mass to visceral fat area is closely related to the pathophysiology of NAFLD in Korea11 and Japan.12 Previous studies in China have reported the association of android fat and/or skeletal muscle with NAFLD.13 14 However, it remains unclear whether the fat-to-muscle ratio (FMR), a novel assessment of the combined effects of fat and skeletal muscle mass, is associated with the risk of NAFLD. Recently, FMR has been used as a new indicator of IR and cardiometabolic disease.15 Increased fat mass and decreased muscle mass lead to abnormalities in lipid and glucose metabolism, exacerbating IR, which mediates the onset and progression of NAFLD.

Therefore, the aim of our study was to investigate the association between FMR and NAFLD. In addition, the association of FMR with the risk of liver fibrosis and atherosclerotic cardiovascular disease (ASCVD) in patients with NAFLD needs further investigation. In summary, this study was conducted to evaluate the association between FMR and the risk of NAFLD, liver fibrosis and ASCVD.

Methods

Study population

Participants in this study were those who underwent a medical health examination at Wuhan Union Hospital between January 2020 and November 2021. Clinical data were retrospectively collected from the electronic medical record by two members of the research group and verified by the other two members. A total of 1830 people aged 40–79 years who voluntarily underwent body composition analysis and liver ultrasound as part of a health examination were consecutively enrolled. During the health examination, each participant completed a questionnaire collecting self-reported data on sex, age, smoking, drinking, and medical and medication history. For smoking history, participants were defined as either current smokers or ex-smokers. For drinking history, participants were defined as either current alcohol drinkers or ex-drinkers. Individuals who reported excessive alcohol consumption, defined as >210 g/week for men and >140 g/week for women, and a history of known liver disease, including viral, autoimmune and drug-induced liver disease, were excluded from the analyses. Individuals with a diagnosis of acute illness, renal insufficiency (estimated glomerular filtration rate <60 mL/min/1.73m2), or active cancer (defined as self-reported history of cancer diagnosed or treated in the past 6 months), those using oral or injectable steroids, and those with missing biochemical measurements or medical history interview records were also excluded. Ultimately, 1592 participants were included in this study. Participants were diagnosed with fatty liver based on the results of conventional abdominal B-mode ultrasound scans (Philips IU22, Philips Healthcare, Inc. N.V.) performed by trained technicians. Participants who had fatty liver without any of the above liver comorbidities were defined as having NAFLD.

Anthropometric and laboratory measurements

Anthropometric measurements were taken by a professionally trained technician in a dedicated body composition analysis room in the hospital. Participants were asked to fast, wear light clothing and stand barefoot on the machine. Body weight, fat mass and muscle mass were measured using a bioelectrical impedance analyser (BIA) (Tsinghua Tongfan, BCA-2A, China), which is based on the theory of body impedance measurement and uses a segmented multi-frequency body impedance measurement model for body composition analysis. The BIA was periodically calibrated by the manufacturer. Height was also measured, and body mass index (BMI) was calculated by dividing the participant’s weight by their height squared. Blood pressure (BP) was measured with an electronic sphygmomanometer (Panasonic, EW3106, China) while the participants were seated and after at least 10 min of rest. BP measurements were repeated at 5 min intervals, and the average of the two measurements was used. Laboratory measurements were checked in the central clinical laboratory of the hospital. After ≥8 hours of fasting, venous blood samples were taken for the determination of biochemical parameters, including platelet count (PLT), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), alanine aminotransferase (ALT), aspartate aminotransferase (AST), uric acid (UA) and fasting blood glucose (FBG). Blood count (PLT) was measured using a haematology analyser from BeckmanCoulter. Biochemical parameters were measured using a Beckman AU5800 fully automated biochemical analyser, with TG measured by the glycerol phosphate oxidase method, total cholesterol by the oxidase method, LDL-C by the selective solubilisation method and HDL-C by the chemically modified enzyme method. FMR was calculated as total body fat mass divided by total body muscle mass and was categorised into tertiles from lowest (T1) to highest (T3). Diabetes was defined as the participant’s self-reported history of diabetes or the use of glucose-lowering medication. Similarly, hypertension was defined as the participant’s self-reported history of hypertension or the use of oral antihypertensive medication.

Definition of risk of liver fibrosis and ASCVD

The burden of liver fibrosis was assessed using the fibrosis-4 index (FIB-4) score, a validated non-invasive index.16 The FIB-4 cut-off points were defined as <1.30 (low risk), 1.30–2.67 (moderate risk) and >2.67 (high risk) to predict the likelihood of NAFLD-related liver fibrosis. As the number of participants classified as high risk (FIB-4>2.67) for NAFLD-associated liver fibrosis was too small to obtain a reliable estimate, we combined the participants at moderate and high risk for liver fibrosis into one group. The ASCVD risk score was calculated using the 10-year ASCVD risk prediction model from the 2013 American College of Cardiology/American Heart Association guidelines.17 High ASCVD risk was defined as an ASCVD score >10%, and low ASCVD risk was defined as an ASCVD score ≤10%.

Statistical analysis

First, all participants were divided into groups with and without NAFLD, and descriptive statistics were performed using frequencies (percentages) for categorical variables, means plus or minus SD for continuous variables with normal distribution, and medians (lower quartile, upper quartile) for continuous variables with non-normal distribution between the two groups. χ2 test for categorical variables, independent samples t-test for normally distributed continuous variables and Mann-Whitney U test for non-normally distributed continuous variables were used for comparisons between two groups. Binary logistic regression was used to analyse the association between FMR and NAFLD. Model 1 was unadjusted, model 2 was adjusted for sex, age, tobacco use, alcohol use, hypertension, diabetes and model 3 was further adjusted for ALT, AST, UA, TG, HDL-C and LDL-C. A receiver operating characteristic (ROC) curve for the detection of NAFLD by FMR was constructed according to sex. Gender-specific FMR cut-offs, sensitivity and specificity were calculated using Youden’s index. Second, patients with NAFLD were divided into three groups according to FMR tertiles. Descriptive statistics were as above. χ2 test for categorical variables with α-splitting method for further two-way comparisons, one-way analysis of variance for normally distributed continuous variables or Kruskal-Wallis test for non-normally distributed continuous variables with Bonferroni post hoc comparison analysis for further two-way comparisons were used to compare between FMR tertiles. Binary logistic regression was used to analyse the association of FMR with moderate/high risk of liver fibrosis and high risk of ASCVD. Model 1 was unadjusted, model 2 was adjusted for sex, tobacco use, alcohol use, hypertension, diabetes and model 3 was further adjusted for ALT, AST, UA, TG, HDL-C, LDL-C, PLT and BMI. ORs along with 95% CIs and p values are reported. Statistical significance was defined as a two-tailed p value <0.05. All analyses were performed using SPSS V.25 (IBM, Armonk, NY, USA).

Patient and public involvement

None.

Results

Clinical characteristics

A total of 1592 participants, of whom 27.9% were female, were included in this analysis. The mean age was 56.8±9.1 years. The percentage of participants with NAFLD was 61.1% (95% CI 0.587 to 0.635). Baseline characteristics of individuals with (n=973) and without (n=619) NAFLD are shown in table 1. NAFLD was more common in men, smokers and drinkers, and also more common in those with dyslipidaemia, hypertension and diabetes. Compared with those without NAFLD, those with NAFLD had more adverse metabolic factors, such as higher BMI, TC, TG, UA, FBG, systolic BP (SBP), diastolic BP (DBP) and ASCVD risk scores and lower HDL-C. Meanwhile, individuals with NAFLD had higher ALT and AST, indicating poorer liver function. In addition, individuals with NAFLD had a significantly higher FMR than those without NAFLD (p<0.001). Similarly, the prevalence of NAFLD was significantly higher in FMR tertile 3 (75.6%) than in tertile 2 (61.2%) and tertile 1 (46.4%) (p<0.001).

Table 1

Baseline characteristics of individuals with and without NAFLD

Association between FMR and NAFLD

Table 2 shows the association between FMR and NAFLD. There were significant differences in the risk of NAFLD according to the FMR tertiles, as a higher FMR was associated with a higher risk of NAFLD in the unadjusted model. After adjustment for covariates such as sex, age, tobacco use, alcohol use, hypertension and diabetes, the risk of NAFLD continued to increase progressively with increasing FMR tertiles. The results remained unchanged after further adjustment for ALT, AST, UA, TC, TG, HDL-C and LDL-C. Using FMR tertile 1 as a reference, tertile 2 (OR=1.49, 95% CI 1.13 to 1.97) and tertile 3 (OR=2.85, 95% CI 2.08 to 3.90) were significantly associated with NAFLD, even after adjustment for conventional metabolic risk factors.

Table 2

Association between fat-to-muscle ratio and NAFLD

FMR cut-off for the detection of NAFLD

Figure 1 shows the ROC curves, cut-off values and area under the ROC curve (AUC) values. To account for sex differences in body composition, the ROC curves were plotted separately for women and men. The cut-off value was higher in women than in men (0.5583 vs 0.3244). The AUC was higher in women (0.703, 95% CI 0.655 to 0.751) than in men (0.638, 95% CI 0.604 to 0.672), and both were statistically significant (p<0.001). Sensitivity, specificity, positive predictive value and negative predictive value are shown in online supplemental appendix table 1.

Figure 1

Receiver operating characteristic (ROC) curves of the fat-to-muscle ratio for the detection of non-alcoholic fatty liver disease according to sex.

Baseline characteristics of patients with NAFLD according to FMR tertiles

Table 3 shows the general characteristics of patients with NAFLD according to FMR tertiles. With increasing FMR, BMI gradually increased and age also tended to gradually increase. Compared with patients in tertile 1 and tertile 2, patients in tertile 3 had significantly higher ALT, AST, FBG, SBP, FIB-4 score and ASCVD risk score. However, there were no significant differences in lipids, UA, DBP, tobacco use or alcohol use between patients in tertile 1 and tertile 3. In addition, among the tertiles, patients in tertile 3 had the highest prevalence of metabolic diseases, including hypertension and diabetes. Meanwhile, the proportion of patients at moderate/high risk of liver fibrosis and at high risk of ASCVD was significantly higher in tertile 3 (48.4 vs 63.0, respectively) than in tertile 1 (37.0 vs 43.1) and tertile 2 (34.3 vs 49.1) (figure 2).

Table 3

Baseline characteristics of patients with NAFLD according to tertiles of fat-to-muscle ratio

Figure 2

Proportion of moderate/high risk of liver fibrosis (A) and high risk of ASCVD (B) according to tertiles of fat-to-muscle ratio in patients with non-alcoholic fatty liver disease. ASCVD, atherosclerotic cardiovascular disease; FIB-4, fibrosis-4 index.

Association of FMR with risk of liver fibrosis and ASCVD

Table 4 shows the association of FMR with the risk of liver fibrosis and ASCVD. Using FMR tertile 1 as a reference, logistic regression analysis showed that tertile 3 (OR=4.48, 95% CI 2.12 to 9.50), but not tertile 2, was significantly correlated with moderate/high risk of liver fibrosis, even after adjustment for multiple confounders such as sex, tobacco use, alcohol use, hypertension, diabetes, ALT, AST, UA, TC, TG, HDL-C, LDL-C, PLT and BMI. Similarly, compared with FMR tertile 1, tertile 2 (OR=2.00, 95% CI 1.25 to 3.22) and tertile 3 (OR=4.63, 95% CI 2.62 to 8.19) were significantly associated with high ASCVD risk, even after adjustment for the above confounders.

Table 4

Association between FMR tertiles and the risk of liver fibrosis and ASCVD in patients with NAFLD

Discussion

Obesity, most commonly assessed by using BMI, has a strong correlation with the prevalence of NAFLD.18 However, Asian individuals are more prone to central fat accumulation despite having a normal BMI and are also more likely to develop NAFLD at a lower BMI than Caucasians.19 In addition, a population-based longitudinal cohort study of 4427 healthy Japanese individuals showed that fatty liver does not develop at the peak of the BMI trajectory, but rather later.20 While these phenomena may be partly explained by genetic factors, they are also relevant to changes in body composition, such as increased fat mass and decreased muscle mass.21 Previous studies have shown that FMR measured by BIA is associated with IR, metabolic syndrome15 and CVD.22 In this study, we found that FMR was significantly positively associated with NAFLD, with a progressive increase in the prevalence of NAFLD from tertile 1 to tertile 3. We also derived the sex-specific optimal cut-off values for FMR to predict the risk of NAFLD and found that the AUC value was higher in women than in men, suggesting that FMR is a better predictor of NAFLD in women than in men. In addition, several studies have shown that the leading cause of death in patients with NAFLD is CVD, followed by extrahepatic malignancies and liver-related complications.23 Therefore, we further investigated the associations of FMR with the risk of NAFLD-related liver fibrosis and ASCVD. We observed that FMR was significantly associated with the risk of liver fibrosis and ASCVD, as NAFLD patients in FMR tertile 3 had a significantly higher risk than those in tertiles 1 and 2.

Similar to our findings, a cross-sectional analysis of 3589 participants from China reported that there was an independent positive association of the android fat ratio and an inverse association of the skeletal muscle index (SMI) with NAFLD defined by the fatty liver index.13 Another study from China conducted by Guo et al, using transient elastography to assess the degree of liver steatosis and stiffness, found that SMI was independently associated with the severity of liver steatosis and fibrosis.14 Shida et al found that NAFLD patients in Japan with a reduced ratio of skeletal muscle mass to visceral fat area had an increased risk of moderate to severe steatosis and advanced fibrosis.12 Lee et al conducted a retrospective cohort study of 4398 initially NAFLD-free subjects from Korea. After 10 years of follow-up, they found that both increased fat mass and decreased appendicular skeletal muscle mass with age were significantly associated with incident NAFLD.24 However, none of these studies examined the risk of CVD in NAFLD patients. Recent guidelines from the American Association for the Study of Liver Diseases and the European Association for the Study of the Liver have highlighted the importance of assessing cardiovascular risk in patients with NAFLD.25 Studies on the association between fat, muscle and ASCVD risk in patients with NAFLD are scarce. A post hoc analysis of 3042 participants from the Attica region of Greece showed that increasing SMI and lower abdominal obesity were independently associated with a lower rate of NAFLD and had an additive effect in determining the 10-year CVD risk in NAFLD patients.26 However, this study used a non-invasive hepatic steatosis index to diagnose fatty liver and an equation to calculate skeletal muscle mass. Our study used the more accurate liver ultrasound to diagnose NAFLD and the well-established BIA to measure body composition. Chun et al, using the Korea National Health and Nutrition Examination Survey (KNHANES) database, reported that sarcopenic subjects with metabolically associated fatty liver disease (MAFLD) had a 7.9-fold increased odds of ASCVD risk compared with subjects without MAFLD.27 Another analysis from the KNHANES showed that NAFLD patients with significant liver fibrosis and sarcopenia were significantly associated with a higher risk of ASCVD.28 However, sarcopenia was uncommon in NAFLD patients,29 and while the former is usually associated with frailty and malnutrition, the latter is usually associated with obesity and overnutrition,30 and skeletal muscle mass increases modestly with weight gain.31 Therefore, it is more appropriate to use FMR for risk stratification in NAFLD patients, which may facilitate earlier identification of ASCVD risk.

Furthermore, in our study, the association of FMR with the risk of liver fibrosis and ASCVD remained robust after adjustment for multiple covariates, including conventional cardiometabolic risk factors, and even after correction for BMI. Interestingly, although BMI is strongly associated with incident NAFLD, it does not appear to be associated with the prognosis of NAFLD. A retrospective cohort study of 646 patients with biopsy-proven NAFLD reported that patients with lean NAFLD had a higher risk of severe liver disease than those with a higher BMI during a median follow-up of 19.9 years.32 A study of 4786 individuals in Korea found that participants with lean NAFLD had a significantly higher risk of ASCVD than those with obese NAFLD.33 Furthermore, in a prospective cohort of Western individuals with NAFLD, Arvind et al showed that the risk of incident CVD was comparable in subjects with and without obesity.34 As such, BMI should not be the sole criterion for screening for NAFLD, nor is it suitable for predicting hepatic or extrahepatic complications of NAFLD. In contrast, a retrospective cohort study performed from Japan suggested that a reduced FMR was associated with improvement in liver stiffness, but reduced BMI was not.35 Therefore, FMR may be a useful adjunct to conventional measures of obesity in assessing the risk of incident and progression of NAFLD.

Although the underlying mechanisms are not fully elucidated, IR and chronic inflammation are thought to be the pathogenesis linking fat, muscle and NAFLD. As target organs for insulin action, enlarged adipose tissue and reduced skeletal muscle promote peripheral IR, which increases the hepatic uptake of free fatty acids released by adipose tissue lipolysis, impairs the suppression of gluconeogenesis and ultimately leads to hepatic steatosis.36 37 In addition, adipose tissue and skeletal muscle are also important endocrine organs, with the former secreting deleterious pro-inflammatory factors and the latter secreting beneficial myokines. The expansion of adipose tissue and the loss of skeletal muscle are involved in the imbalance of adipokines and myokines and thus mediate the progression of NAFLD.12 In addition, steatosis and fibrosis in the liver release a variety of inflammatory cytokines and proatherosclerotic mediators that contribute to ASCVD.23

Despite the above findings, our study has several limitations. First, due to the cross-sectional design of the study, we cannot infer a causal relationship between FMR and NAFLD. A longitudinal study is needed in the future. Second, in this study, body composition was measured by BIA rather than dual-energy X-ray, which is more accurate, but is difficult to use in clinical practice because of its high cost and harmful radiation. Third, we used liver ultrasound and the liver fibrosis index in our analysis instead of liver biopsy, which is invasive, risky and not feasible in large population screening. Fourth, data on IR, adipokines and myokines are not available, so the underlying mechanisms cannot be elucidated. Finally, this is a single-centre retrospective study, which limits the generalisability of the findings to other populations.

Conclusions

The present study showed that FMR was an independent risk factor for NAFLD in a middle-aged and elderly adult population in China, and it was also a predictor of liver fibrosis burden and ASCVD risk in patients with NAFLD. Careful evaluation of FMR is needed in the follow-up of weight management in patients with NAFLD, as weight loss may improve NAFLD but may worsen skeletal muscle loss, which also contributes to the risk of liver fibrosis and ASCVD. Reducing fat mass and increasing muscle mass through exercise, rather than diet alone, is an important strategy for the prevention and management of NAFLD.

Data availability statement

Data are available upon reasonable request. Extra data can be accessed via the Dryad data repository at http://datadryad.org/ with the doi:10.5061/dryad.7d7wm3809.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by the Institutional Review Board of Tongji Medical College, Huazhong University of Science and Technology (S155). Informed consent was not required because all medical data were retrospectively reviewed and analysed anonymously.

Acknowledgments

We would like to acknowledge A/Prof Anna Wong Shee, Grampians Health, for their support in the design of the study and review of the manuscript.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors Study concept and design: FY and WP. Data acquisition: FY, GN and NZ. Data analysis: FY and MZ. Data interpretation: MZ and WP. All authors contributed to drafting, revising and approving the final manuscript. WP is responsible for the overall content as guarantor. The guarantor takes full responsibility for the finished work and the conduct of the study, had access to the data and controlled the decision to publish.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.