Insulin sensitisers in the treatment of non-alcoholic fatty liver disease: a systematic review

The accumulation of liver fat

Hepatic steatosis (accumulation of fat within liver cells) is the first stage of NAFLD. There is no single mechanism leading to hepatic steatosis, but rather the combination of a number of pathologies that ultimately disrupt normal lipid [fat-rich products, mainly triglycerides (TGs) – fatty acid molecules] movement through the liver cell and cause lipid accumulation. The fat in the liver can be traced to three sources – dietary intake, de novo synthesis and circulating non-esterified fatty acids (NEFAs) derived from body fat stores. 5 It is known that high-fat diets can lead to hepatic steatosis, but it appears that NEFAs are the main source of liver fat, with 60% in individuals with NAFLD who have a normal fat-containing diet. 12

The body’s resistance to the effects of the hormone insulin (required for the uptake of glucose – the main sugar derived from dietary carbohydrate by liver, fat and muscle cells) is thought to play a key role in increasing levels of NEFAs, particularly the insulin resistance shown by fat and muscle cells (Figure 1). Once lipid starts to accumulate in the liver this can, in fact, worsen the body’s insulin resistance, reducing the beneficial effects of insulin, leading to a vicious cycle. 13–15

FIGURE 1.

The effects and potential clinical indicators of accumulating liver fat. The presence of ectopic fat in the liver cell leads to hepatic insulin resistance following the accumulation of intracellular lipid by-products, which leads to disturbed glucose metabolism. Gluconeogenesis is the production of glucose from protein and fat breakdown, and glycogen is the form of complex carbohydrate that glucose is stored as. BMI, body mass index; FPG, fasting plasma glucose. Adapted from Preiss and Sattar. 5

Inflammation of the fatty liver – non-alcoholic steatohepatitis

McCullough16 has reviewed the pathophysiology of NASH. Hepatic steatosis is considered to have a benign course in most cases. 17 This may be because of associated counter-regulatory protective mechanisms, which means that liver tissue changes (other than steatosis) and liver function may remain normal. NASH represents the stage when the fatty liver starts to show inflammatory change. It is the development of NASH and progression to fibrosis and cirrhosis that is responsible for the liver-specific morbidity and mortality of NAFLD.

It is unclear why only approximately 25–30% of patients with simple steatosis develop NASH. 16 The counter-regulatory processes occurring as a means to maintain liver cell (hepatocyte) health forms the basis of the older ‘two-hit’ theory in the development of NASH. A newer theory suggests that NASH occurs by liver fat directly causing inflammation.

In the two-hit theory, the first hit refers to the accumulation of liver fat. It has been proposed the hepatocytes act, in addition, as a reservoir of toxic agents and are most susceptible to a ‘second hit’ – oxidative stress18,19 – caused by endogenous compounds within liver cells and by environmental toxins. 20,21 The cells in the body constantly react with inhaled oxygen, producing energy – oxidation. As a consequence of this activity, highly reactive molecules (free radicals) are produced. Free radicals interact with other molecules within cells. This can cause damage – oxidative stress – to proteins, membranes and genes. Patients with NASH have increased levels of oxidative stress when compared with patients with steatosis alone. 22,23

In addition to endogenous toxins, three factors have been proposed as potential causative mechanisms for producing this oxidative stress: increased insulin levels, lipid peroxidation and liver iron content (Figure 2).

FIGURE 2.

The two-hit theory of NASH. This involves insulin resistance, which causes hepatic steatosis, a process enhanced by obesity and/or T2DM. Once developed, hepatic steatosis may remain in a benign state or progress to NASH via the mechanisms discussed. Reproduced with permission from AJ McCullough. Pathophysiology of nonalcoholic steatohepatitis. J Clin Gastroenterol 2006;40(Suppl. 1):17–29. 16

Insulin may injure the liver both directly and indirectly. 24,25 The indirect effects are related to the hyperinsulinaemic state increasing liver fat accumulation, as described earlier. The direct effect may be due to insulin’s ability to generate oxidative stress. 25 It also appears to have direct effects by stimulating scar tissue formation, especially in the presence of increased glucose levels. 26 This may explain the observation that NAFLD patients with T2DM have a particularly poor prognosis. 17,27 Insulin may be directly involved in causing stress on parts of the liver cell that lead to cell death (apoptosis). This, in itself, may exacerbate insulin resistance. 28

Increased lipid peroxidation (breakdown of NEFAs, causing oxidative stress) has been demonstrated in patients with NAFLD. 22,23,29–31 As previously described, patients with NAFLD have increased breakdown of fat stores and increased delivery of NEFAs to the liver. 22,32 The products of NEFA oxidation are capable of generating oxidative stress and subsequent lipid peroxidation, setting up a vicious cycle.

In response to oxidative stress, there is usually an increased synthesis of antioxidants. However, NAFLD patients have decreased antioxidants (glutathione). 33 Therefore, patients with NAFLD have an impaired ability to produce sufficient antioxidants. A deficiency in antioxidants is also supported by a preliminary report that demonstrated that betaine (a naturally occurring antioxidant) improved the microscopic images of liver tissue and liver function tests (LFTs) in patients with NAFLD. 34

The role of iron as a pro-oxidant in NASH is unclear. In McCullough’s review,16 there is mention of patients with NAFLD having increased ferritin levels, and a relationship between hepatic iron and insulin resistance. However, there is acknowledgement that there is no strong evidence associating iron overload with NAFLD. Therefore, for now, iron is likely to be important in only a minority of patients with NASH and more research is required on iron as a pathophysiological factor.

There is also good evidence to support a one-hit theory of liver fat causing chronic inflammation directly, leading to direct NAFLD progression. 5 A key player in steatosis formation appears to be nuclear factor-κB (NF-κB). This is a transcription factor – an intracellular protein required for the transcription of deoxyribonucleic acid to form proteins. In animal models it has been shown that a high-fat diet with resultant hepatic steatosis leads to increased NF-κB signalling in the liver. 35 This then induces the production of chemicals involved with causing inflammation, which may play a role in the progression of NAFLD. The transcription factor also leads to activation of specific cells (Kupffer cells and macrophages) within liver tissue, which are known to cause further damage to liver tissue. In the same study35 there is evidence that inflammation, in the form of isolated increased NF-κB expression in rat liver, can lead directly to insulin resistance. 35

Another factor that may be involved in the process is adiponectin, a polypeptide produced in adipocytes (fat cells) and which may be protective via an insulin-sensitising effect in the liver. Polyzos et al. 36 reviewed the evidence on the role of adiponectin and noted that it was reduced in NAFLD and NASH, and increased by most interventions which improved NAFLD, including weight loss, bariatric surgery and pioglitazone, but not metformin. However, formal interventions and links to adiponectin genotypes would be required to prove a causal relationship.

The advanced stages – fibrosis and cirrhosis

As NAFLD progresses, more advanced forms can occur – fibrosis and cirrhosis. Liver fibrosis is the formation of excess fibrous tissue in the liver37 and is a reparative or reactive process, as a result of NASH,38 or, in a few cases, direct progression from simple steatosis. 39 Liver fibrosis can lead to loss of function.

Cirrhosis of the liver is advanced fibrosis associated with regenerative nodules (an attempt at repair). 40 Cirrhosis is associated with variable and usually irreversible loss of liver function. When liver function is minimally or not significantly compromised clinically, it is often termed ‘compensated cirrhosis’; however, when there is clinical evidence that the cirrhotic liver is unable to function properly, it is termed ‘decompensated cirrhosis’. Whereas decompensation can be reversible if due to an acute insult (e.g. infection), decompensation is often progressive, resulting in liver failure and death. Hui et al. 41 in their prospective cohort study followed up 23 patients with NASH-associated cirrhosis for a mean duration of 84 months (range 5–177 months) and found that 9 out of 23 cases developed liver-related morbidity (eight developed ascites and/or encephalopathy, one developed variceal bleeding). The authors then found that probability of complication-free survival was 83%, 77% and 48% at 1, 3 and 10 years, respectively, and the cumulative probability of overall survival at 1, 3 and 10 years was 95%, 90% and 84%, respectively. 41

International data

As previously mentioned, screening studies using serum liver tests and ultrasonography (USG) suggest that the prevalence of NAFLD ranges from 17% to 33% in the general population of the Western world. 50 Imaging using magnetic resonance spectroscopy (MRS) gives NAFLD a prevalence of 34%. 51 The prevalence of NASH is less well known, as a liver biopsy is required to confirm the diagnosis. Estimates of the prevalence are in the range of approximately 3% in the general population and higher among obese persons. 52 The average age at diagnosis lies between 45 and 55 years,45 and has a slight female preponderance. 45

Natural history

Good long-term data (including UK data) on the natural history of NAFLD from simple steatosis to more advanced stages are lacking, for a number of reasons. NAFLD has only recently been characterised in detail. There are few long-term follow-up studies of well-defined patient cohorts, and follow-up liver biopsies have been performed in only a limited number of patients. The diagnostic method for fatty liver – blood tests ± ultrasound ± liver biopsy (the current gold standard)50 – is not uniform, and studies using ultrasound to diagnose steatosis will give higher numbers than liver biopsy. Furthermore, it is possible that the long-term complications of NAFLD may be under-recognised and under-reported, as the characteristic features of fatty liver, such as steatosis, may disappear in the late stages of the disease, leading to a picture of ‘bland’ cirrhosis, frequently described as ‘cryptogenic’, rather than NAFLD-related cirrhosis, which is now recognised as the most common cause of cryptogenic cirrhosis. 53 Finally, most studies to date that have studied the natural history of NAFLD have been retrospective analyses (e.g. clinical follow-up from cohort studies) or case series in which selected patients with a diagnosis of NASH underwent subsequent liver biopsies. 6

Despite these limitations, the following diagrams indicate the current thinking on the prevalence and progression of NAFLD (Figures 3 and 4).

FIGURE 3.

The natural history of NAFLD. Adapted from Preiss and Sattar. 5 a, Matteoni et al. ,17 Day;39 b, Fassio et al. ,54 McCullough. 16

FIGURE 4.

The natural history and clinical outcomes of NASH. OLTX, orthotopic liver transplantation. Reproduced with permission from AJ McCullough. Pathophysiology of nonalcoholic steatohepatitis. J Clin Gastroenterol 2006;40(Suppl. 1):17–29. 16

Hence, it is difficult to determine what proportion of an unselected population will develop NAFLD-related cirrhosis and subsequent HCC. One study estimated that 7% of cases of HCC were related to underlying NAFLD or cryptogenic cirrhosis but such data are, at best, approximates. 9

UK data

The UK data on NAFLD are limited, with no nationwide data. The incidence of NAFLD in a hepatology catchment area in England of 200,000 residents, based on referrals to a secondary care setting, was calculated at 29 per 100,000 population. 55 This was subdivided into 23.5 per 100,000 population for non-cirrhotic NAFLD and 5.5 per 100,000 population for cirrhotic NAFLD. There are no satisfactory prevalence data.

Associations with type 2 diabetes mellitus and obesity

Numerous studies have reported the increased prevalence of NAFLD in individuals with T2DM,56 and increased risk and severity of NAFLD in T2DM. 57–59 In 939 randomly selected people with T2DM in Edinburgh, ultrasound-detected steatosis was present in 73.9% of subjects. 60 A recent study in the USA reported the prevalence of ultrasound-determined NAFLD to be 69.5%. 56 The prevalence of NAFLD in obese individuals (ultrasound determined) has been estimated to be high as 80%. 50 One thing is certain – as the incidence and prevalence of obesity and T2DM increase,61 the incidence and prevalence of fatty liver, and hence NASH and more severe forms of NAFLD, are also likely to increase.

Cardiovascular risk

In addition to the organ-specific related morbidity and mortality of NAFLD, NAFLD has also been linked with increased cardiovascular risk, largely through the components of the metabolic syndrome. A detailed review in 2007 on NAFLD and cardiovascular risk62 showed that this increased risk was related to the presence of known cardiovascular risk factors, several of which (insulin resistance, obesity and dyslipidaemia/hypertriglyceridaemia) are also well associated with NAFLD. 63

Similar conclusions were reached in a more recent review by Ghouri et al. ,64 who concluded that the presence of NAFLD was an indication for screening for T2DM, but that it did not add useful data on CVD risk compared with traditional risk factors.

Hence, it appears that NAFLD itself is not an independent contributor to CVD risk, but that it is associated with adverse risk factors.

A more recent review by Targher et al. 65 also addressed the question of whether or not NAFLD increased the risk of cardiovascular disease, independent from its association with traditional risk factors. They concluded that:

Although additional research is required to draw a definite conclusion, these observations raise the possibility that NAFLD – especially its necroinflammatory variant, NASH – not only is a marker of cardiovascular disease but may also be involved in its pathogenesis. This process may occur through the systemic release of pro-atherogenic mediators from the steatotic and inflamed liver or through the contribution of NAFLD itself to insulin resistance and atherogenic dyslipidaemia.

One key issue noted by Targher et al. 65 is that cardiovascular disease is a greater threat to people with NAFLD than liver disease.

Associations with type 1 diabetes

A recent study66 has reported a high prevalence of NAFLD in 202 patients with type 1 diabetes in Italy. NAFLD was diagnosed by history and liver ultrasound. Over half of the group were classed as having NAFLD. Those who did were older, had suffered from diabetes longer and had higher body mass indices (BMIs) than those who did not.

Lifestyle changes

Reduced calorie intake and increased physical activity are viewed as logical methods to reduce liver fat content. Two studies have shown that a reduction in weight by 10% significantly reduced elevated LFTs compared with subjects with NAFLD who did not lose as much weight. 71,72 Park et al. 72 from Republic of Korea reported a marked reduction in liver enzymes (AST and ALT) in those who lost weight, but not in those who did not. Ueno et al. 73 from Japan reported that effects of intensive lifestyle modification over a 3-month period in overweight patients with NAFLD produced an improvement in the microscopic changes in liver tissue, but such changes did not attain significance. 73 However, a reduction of three units in BMI normalised previously elevated AST and ALT levels.

Weight loss drugs

Published pilot data on orlistat, a weight loss drug that reduces fat absorption, have shown improvements in LFTs, ultrasound findings and microscopic changes in liver tissue. 74,75 A further small study76 compared orlistat with another drug – sibutramine, an appetite suppressant – and the results showed an improvement in LFTs and reduced liver fat on ultrasound in both groups.

In a detailed review of the effects of lifestyle modification with and without the use of weight loss drugs in patients with steatosis and NASH, Harrison and Day77 concluded that a weight loss of 10% of body weight not only improved the biochemical measures seen in NAFLD, but also improved the histological changes seen in NASH. They also concluded that the evidence from a number of studies suggested that physical activity alone, or with only modest (3%) weight loss, was also effective in improving insulin sensitivity.

More recently, a high-quality systematic review by Musso et al. 78 reviewed all treatments for NAFLD. The authors also concluded that weight loss improved or reversed NASH, but that it appeared from two randomised controlled trials (RCTs) that weight loss had to be at least 7% or 9% for histological features to measurably improve. Musso et al. 78 came to the same conclusion as Harrison and Day77 (but based on different studies) – that the evidence suggested that physical activity improved NAFLD independently of weight loss. These data, in turn, concur with evidence for an independent association of higher activity levels, albeit assessed by questionnaire, with lower GGT levels in a cohort study of British women. 79

Drugs that control blood glucose levels

Trials using drugs that improve the body’s sensitivity to the effects of insulin have been performed in individuals with NAFLD. Two main drugs – metformin and thiazolidinediones (TZDs) (glitazones) – have been investigated. Metformin reduces glucose production in the liver and improves the uptake of circulating glucose in the blood by fat and muscle, whereas the glitazones improve the uptake of circulating glucose in the blood and are also believed to redistribute fat away from ectopic sources (particularly the liver) to subcutaneous areas, often with an overall increase in weight (substantial in some individuals). 80

The glitazones have adverse effects, including oedema, higher risk of fracture and in addition the development and worsening of heart failure;81,82 and rosiglitazone moderately increases cardiovascular risk83,84 and has recently been suspended in Europe.

The evidence on metformin and the glitazones is reviewed in Chapter 2.

Non-alcoholic fatty liver disease and other drugs

Patients with NAFLD are likely to be considered for lipid-lowering statin therapy owing to their elevated lipid levels and increased cardiovascular risk. Statin therapy in NAFLD certainly appears safe and should not be avoided because of mildly abnormal LFTs. 85 There is also some evidence of improvement in liver histology on statin therapy from a small placebo-controlled study that was recently published. 86 Other drugs studied have included vitamin E and fibrates. Vitamin E therapy has produced variable results87,88 and fibrate therapy has not shown benefit thus far. 89 In the only placebo-controlled studies so far, combination therapy with the bile component ursodeoxycholic acid (UDCA) with vitamin E for 2 years resulted in a significant reduction in steatosis. 90 UDCA therapy alone did not improve changes in liver tissue, although there was an improvement in LFTs. In two small studies, treatment with the angiotensin II blocker, losartan, also led to improvements in liver histology. 91,92

A review of all drugs for NAFLD was not in our remit, but we note the findings of the recent review by Musso et al. ,78 who found that:

• There is some evidence of benefit from treatment with polyunsaturated fatty acids. Trials are in progress.

• Fibrates had no effect.

• The evidence on statins was sparse, but there was some evidence of benefit from atorvastatin.

• The lipid-lowering drug probucol lowered ALT, but also high-density lipoprotein (HDL), which might increase cardiovascular risk.

• UDCA had little or no effect.

• There were mixed results with antioxidants.

• Pentoxifylline appeared to be of benefit; two RCTs are under way.

• Telmisartan (an angiotensin receptor blocker) appeared to reduce fibrosis, the only drug to do that. It improved steatosis, ballooning and lobular inflammation. It led to a reduction in insulin resistance, and TG and total cholesterol (TC) levels. It is noteworthy that the effects of valsartan were less despite similar blood pressure effects. The liver effects of telmisartan may be related to its peroxisome proliferators-activated receptor (PPAR) gamma activity.

• There was some evidence of benefit for l-carnitine.

• Orlistat added to diet resulted in little difference – weight loss of 8% instead of 6% on diet alone.

Surgery

Recently, data on surgery for morbid obesity with 18- to 24-month follow-up have been published. There are two main types of surgery for obesity. One procedure, gastric bypass surgery, in essence involves reducing the size of the stomach, by bypassing a large section of the stomach and connecting it to the small intestine, and this can be done laparoscopically (keyhole surgery). The second procedure is adjustable gastric banding, and involves placing an adjustable band over the top part of the stomach, creating a pouch that reduces the size of the stomach, and is also done laparoscopically. Studies have shown improvement of NAFLD staging or even complete disease resolution following surgery. Mummadi et al. 93 carried out a systematic review to explore effectiveness of bariatric surgery in patients with NAFLD and found that the procedure was safe, with resolution of steatosis in 91.6% of patients, improvement of steatohepatitis and fibrosis in 81.3% and 65.5% of patients, respectively, and resolution of NASH in 69.5%. 93

The effects of bypass surgery and concomitant weight loss (mean loss 50 kg), evaluated by liver biopsy, found considerable improvements in the prevalence of steatosis (90% preoperatively to 2.9% post-operatively), hepatocellular ballooning (swelling of liver cells: 58.9% preoperatively to 0% post-operatively) and fibrosis (50% down to 25%). 94 In a similar study,95 there were similarly impressive reductions in steatosis, fibrosis and hepatocellular ballooning. In this latter study on a group of 18 patients, NASH resolved in 84% of patients and steatosis in 75%. Laparoscopic adjustable gastric banding with resultant weight loss has also led to promising improvements in liver histology. 96,97 As obesity surgery is relatively novel, long-term outcome data are still limited, but should become more available as more of these procedures are performed.

Musso et al. 78 concluded that the RCTs of surgery had insufficient follow-up as yet, noting that liver enzymes fluctuated and did not always correlate with fibrosis and NASH. However, a review of bariatric surgery by Kushner and Noble98 noted that two studies had shown resolution or improvement in NASH and fibrosis after bariatric surgery.

Criteria for considering studies for this review

Search strategy

Comprehensive systematic searches of electronic databases were performed in order to retrieve relevant papers.

Searches were conducted in the following sources to identify both published studies and meeting abstracts:

• MEDLINE, 1950 to June 2010; EMBASE, 1980 to June 2010; Science Citation Index – Expanded, June 2010 (limited to meeting abstracts only); Conference Proceedings Citation Index – Science, June 2010; The Cochrane Library, 2005–2010; and authors’ reference lists. Websites such as ‘ClinicalTrials.gov’ were systematically searched to find any ongoing trials.

The following MEDLINE search strategy (Ovid) was adapted for use with the other databases:

1. fatty liver/

2. liver.tw.

3. (non-alcoholic or non alcoholic or nonalcoholic).tw.

4. (fatty or steato*).tw.

5. (NAFLD or NAFLD or NASH).tw.

6. 4 and 3 and 2

7. 6 or 1 or 5

8. exp Metformin/

9. exp Thiazolidinediones/

10. (insulin sensit* or metformin or pioglitazone or rosiglitazone or thiazolidinedione* or glitazone*).tw.

11. 18 or 9 or 10

12. 7 and 11.

All of the relevant systematic reviews were searched for additional studies. No attempt was made to find unpublished studies. No language restriction was applied to the search strategy.

Details of the electronic search strategies used for the review of clinical effectiveness are given in Appendix 1.

Identification of studies

Abstracts and titles retrieved by the search strategy were assessed independently by two researchers and screened for inclusion and exclusion. Full texts of studies considered possible inclusions were obtained and each was examined by two reviewers independently. Any discrepancies between the two were resolved by discussions and with involvement of a third reviewer when necessary. The papers that did not meet the inclusion criteria were excluded.

Data extraction strategy

Two reviewers independently extracted data in a specially designed form, and data regarding study design and characteristics, details of the intervention and patient characteristics and outcomes were recorded in the form for each study. Differences in data extraction were resolved by discussion, referring back to the original papers and with involvement of a third reviewer when necessary.

Quality assessment strategy

To assess the quality of the RCTs, the following criteria were used: (1) method and description of randomisation; (2) allocation concealment; (3) blinding; (4) intention-to-treat (ITT) analysis; (5) percentage who completed the trial; (6) power calculation; and (7) similarity of group participants at baseline.

Best practice for each of the criteria would be as follows:

• randomisation random assignment generated by computer

• concealment of allocation those at point of implementing random allocation to treatment do not know to what the next patient will be allocated

• blinding those assessing outcomes (e.g. the pathologist looking at biopsies) should not know which treatment patients were on

• intention to treat patients remained on their allocated treatments throughout with no crossover

• loss to follow-up all patients completed the trial with no losses to follow-up

• baseline matching randomisation ensured that prognostically important variables were equally distributed across the arms.

Studies meeting most of these criteria were regarded as high-to-moderate quality.

Analysis

Meta-analyses of the outcomes were not possible, as outcomes were reported incompletely and in a variety of ways. Hence, all of the results are presented in text and tables.

Result of the searches

A total number of 1842 titles and abstracts were retrieved by the searches (Figure 5). The titles and abstracts were screened for inclusion and exclusion; 49 were considered possible inclusions and full texts of these were obtained. Out of these, 34 papers were excluded because of not meeting the inclusion criteria, not reporting outcomes of interest or not being RCTs. Details of the reasons for exclusion are given in Figure 5.

FIGURE 5.

Flow chart of search results.

A total of 15 RCTs (14 full texts and one abstract100) fulfilled the inclusion criteria and were included in the review. Of the 15 studies included, four examined pioglitazone,101–104 eight metformin,87,100,105–110 one rosiglitazone111 and two trials compared the effects of metformin and rosiglitazone. 112,113

A search for ongoing trials or reviews was carried out. An out-of-date Cochrane review on insulin sensitisers by Angelico et al. ,114 in Rome, was found. It excluded people with T2DM, and included only three trials: two of metformin and one of pioglitazone. Four systematic reviews were also found but none included all of the trials now available. A systematic review by Musso et al. 78 also included fully published trials and abstracts that had explored the efficacy of non-pharmacological interventions, UDCA, lipid-lowering drugs, antioxidants, anti-tumour necrosis factor alpha agents (pentoxifylline), anti-hypertensive drugs, endocannabinoid receptor antagonists, l-carnitine and bariatric surgery among patients with NASH/NAFLD.

Descriptions of included studies

The included trials (identified by first author and year) are reviewed in this section. Further details of these can be found in the data extraction tables in Appendix 2. Table 3 gives a summary of all of the included studies.

Quality of included studies

A summary of the quality of the studies is given in Table 4. Full details are given in Appendix 3.

Main results

The following outcomes are summarised in this section:

1. liver histology

2. glycosylated haemoglobin (HbA_1c)

3. fasting plasma glucose (FPG)

4. weight/BMI

5. blood pressure

6. lipid parameters

7. liver biochemistry

8. insulin sensitivity/resistance

9. QoL

10. conversion to diabetes/impaired glucose tolerance/metabolic syndrome.

Details of all outcomes are given in the tables below (see Tables 5–12).

Summary

Fifteen RCTs were included in this review. Four trials examined the effects of pioglitazone, seven examined metformin, one rosiglitazone and three compared metformin against rosiglitazone. The duration of most trials ranged between 6 months and 1 year, and only one trial104 lasted 96 weeks with a 24-week follow-up. The total number of participants analysed in this review was 881 (751 excluding participants who received rosiglitazone in combination with other drugs or rosiglitazone on its own). In all of the pioglitazone trials, the dose of pioglitazone was 30 mg once a day. The dose of metformin varied between 1 and 2.5 or 3 g/day, with most trials using 1.7 g/day. The dose of rosiglitazone ranged between 4 and 8 mg/day. In most trials, the participants in the control group as well as in the intervention group also received other measures including counselling on diet and exercise, use of a hypocaloric diet and lifestyle modifications.

Pioglitazone was found to improve all parameters of liver histology, was better than placebo or diet and exercise or a hypocaloric diet, but no better than vitamin E.

Results with metformin were mixed. In comparison with placebo or diet and exercise, some studies suggested that metformin led to some improvements in steatosis and necroinflammation, but not in fibrosis. However, in most studies the changes seen with metformin were not significantly different from those in the control arms. The Bugianesi trial87 reported clear improvements in histology with metformin, but no biopsies were done in the control arm. The ultrasound trials suggest that metformin led to improved liver echogenicity compared with diet and exercise. Uygun et al. 110 found significant differences with ultrasound (p = 0.038), but not with biopsy.

Rosiglitazone showed significant improvement in hepatic steatosis and very little improvement in any of the other parameters. In head-to-head trials comparing metformin and rosiglitazone, the results varied. Idilman (2008)112 found greater improvements in hepatic steatosis, ballooning and fibrosis with metformin than diet and exercise alone. ‘Rosiglitazone only’ led to a reduction in hepatic steatosis. The overall improvement in NAFLD score from baseline to end was better with either metformin or rosiglitazone than with diet and exercise (Idilman 2008112). Omer (2010)113 found significant improvement in NAFLD score with rosiglitazone and a combination of rosiglitazone and metformin and no significant changes occurred with metformin alone.

Pioglitazone was better than any control groups (placebo, diet or exercise) in terms of improvement of ALT, insulin sensitivity, and FPG, but led to weight gains of 2.5–4.7 kg, and increases in BMI of 0.7–4.7 kg/m². The changes in waist–hip ratio, waist or hip circumference and body fat with pioglitazone were not significant. Metformin led to greater reduction in weight, BMI and waist or hip circumference than rosiglitazone or a combination of the two. Although there were reductions in lipid parameters, ALT and insulin resistance with metformin, the changes were similar to those with placebo or lifestyle modifications or rosiglitazone.

Rosiglitazone was better than metformin in terms of HbA_1c levels, FPG and insulin sensitivity, but the changes in weight were inconclusive, with one trial (Ratziu 2008111) showing better results with rosiglitazone and the other with a combination of rosiglitazone and metformin (Torres 2009100). The reductions in BMI, body fat and waist or hip circumference were not greater with rosiglitazone. The changes in lipid profiles with rosiglitazone were inconclusive. A combination of metformin and rosiglitazone was found to be similar to rosiglitazone alone in terms of ALT reduction and insulin resistance (Torres 2009100).

Strengths and limitations

Study selection

Articles for review were selected, based on whether or not they seemed likely to have data on non-invasive alternatives to liver biopsy, ideally with liver biopsy used as a comparator. Reviews were selected in the first instance.

Quality assessment

The intention was to assess the quality of review articles using recognised appraisal frameworks, for example the Critical Appraisal Skills Programme (CASP)130 or the Scottish Intercollegiate Guidelines Network (SIGN). 131 However, the majority of reviews did not describe the search strategy adopted, their inclusion and exclusion criteria or how included studies were appraised. Therefore, this section aims to simply summarise their findings and conclusions.

Laboratory markers of fibrosis

The guideline recently issued by AISF states that several biochemical indicators of fibrosis have been investigated as means of reducing the number of people requiring liver biopsy. 125 However, AISF describes these as having not yet been rigorously tested and validated, as well as being unavailable within many laboratories. 125 Therefore, AISF states that these ‘do not avoid the need to perform liver biopsy in clinical practice.’125

Myers126 categorised such markers as signifying oxidative stress, inflammation or hepatocyte apoptosis. 126 With respect to the evaluation of markers of oxidative stress, he describes the results as generally being ‘… disappointing and/or inconsistent …’, concluding that ‘… additional studies are necessary before markers of oxidative stress can be used clinically to differentiate simple steatosis from NASH’.

With respect to biomarkers of inflammation, Myers126 states that a marker which has been ‘… validated and is ready for clinical use is not available.’ With respect to hepatocyte apoptosis, Myers discussed research concerning the use of caspase-generated fragments of cytokeratin-18 (CK-18). Although encouraging, Myers126 states the findings require validation ‘… in large-scale, multicentre studies that include more diverse patient populations.’

Clinical scores for fibrosis

Several clinical scoring systems, or fibrosis marker panels, have been developed with the aim of differentiating NASH from NAFLD (e.g. NAFLD fibrosis score, European Liver Fibrosis panel, etc.). 125,126 Some of these are subject to patent. 126

These scoring systems attempt to identify people at risk of fibrosis through measuring indicators such as BMI together with various biochemical parameters. The biochemical parameters can be described as:

• indirect markers of fibrosis in that they ‘… reflect alterations in hepatic function, but do not directly reflect extracellular matrix (ECM) metabolism (e.g. liver biochemistry, platelets)’126

• direct markers of fibrosis in that they ‘… reflect the dynamics of the ECM [extracellular matrix] turnover (e.g. matrix metalloproteinases and their inhibitors, collagens, hyaluronic acid)’. 126

Both indirect and direct markers are subject to limitations. For example, changes in some indirect markers may be observed only in the advanced stages of chronic liver disease. 136 Changes in direct markers may be due to non-hepatic disease. 136

In many cases, the biochemical parameters within fibrosis marker panels are a mixture of direct and indirect markers. 126 Many of these panels were developed for use among patients with chronic hepatitis C infection, but their use in other conditions, including NAFLD, is being explored. 126

Poynard et al. 137 undertook a review of biomarkers of liver fibrosis associated with chronic hepatitis C, chronic hepatitis B, alcoholic liver disease, NAFLD and mixed causes. They identified 14 biomarkers, nine of which were not patented and five that were. Only patented biomarkers were considered in detail within the review. Two of the authors have connections to a company marketing one of the patented biomarkers considered – FibroTest™ (Biopredictive, Paris, France). They found no significant differences between the five patented tests. With respect to performance in relation to the aetiology of fibrosis, the authors stated that most studies had been performed among patients with chronic hepatitis C, with only one test – FibroTest – having ‘… been investigated specifically in the four most frequent chronic liver disease(s)’. Only three of the studies of patented biomarkers included in the review – two involving FibroTest (267 patients) and one involving the European Liver Fibrosis panel (61 patients) – examined the use of biomarkers among individuals with NAFLD. The authors stated that ‘neither biomarkers nor biopsy are sufficient alone to take definitive decision(s) in a given patient …’ and suggested ‘… a moratorium on liver biopsy as a first line procedure while awaiting studies demonstrating biopsy cost-utility vs that of biomarkers’.

Poynard137 and another author138 contributing to the above review (both of whom have connections to the company marketing FibroTest) contributed to a further review focusing on FibroTest as a measure of liver fibrosis. The majority of studies identified assessed the use of the test among patients with chronic hepatitis C. Only two studies examining the use of the test among patients with NAFLD were identified, both of which were included in the earlier review. 137 Similar to the conclusions of the earlier review, the authors stated: ‘FibroTest™ may be an alternative to liver biopsy in the four more common liver diseases – namely, HCV, HBV, NAFLD and ALD. However, neither biomarkers nor biopsy alone are sufficient to allow definitive decisions to be made for a given patient, and all clinical and biological data must be taken into account.’

The AISF guideline125 did not refer to FibroTest, although it referred to several other fibrosis marker panels not included within the Poynard review137 [e.g. BAD (BMI, age, T2DM), BAAT (BMI, age, ALT, TG), HAIR (hypertension, ALT, insulin resistance), BARG (BMI, age, AST/ALT ratio, HbA_1c or glucose), NAFLD Fibrosis Score, BARD (BMI, AST/ALT ratio, T2DM), Clinical Scoring System for Predicting NASH]. Studies describing the use of some of these were published outwith the time period covered by Poynard et al. 137 (February 2001 to July 2007). The AISF guideline states: ‘These indices, particularly the NAFLD fibrosis score, should be considered in evaluating patients for liver biopsy.’125

The majority of studies evaluating the use of fibrosis marker panels have involved people with chronic hepatitis C. Some marker panels have been investigated among patients with NAFLD, although for some the findings of initial studies require to be confirmed by further research. 135,136 At the current time, the relative ability of these panels to differentiate NASH from NAFLD is unclear, and comparisons are difficult as a consequence of studies using a variety of scoring systems and end points in the assessment of fibrosis. 135

At present, the consensus view appears to be that fibrosis panels should be used as an adjunct to liver biopsy rather than as a substitute. Myers126 states:

‘Most experts would agree that the most rational approach is to use these markers as a complement to liver biopsy on a case-by-case basis. It would be unrealistic to expect any index to completely replace liver biopsy which offers a wealth of additional information … It is important to remember that up to a third of patients suspected of having NAFLD have another cause of liver enzyme elevations identified by liver biopsy.’126

Imaging

Ability to compare markers

Interpreting the current data on the effectiveness of non-invasive markers of liver fibrosis is challenging as a consequence of:

• small number of patients recruited in some studies135

• assessment of fibrosis varies across studies135

• the performance of tests being dependent on the prevalence of severe fibrosis in the study population. 136

Area under receiver operating characteristic curve (AUROC) estimates are a commonly used measure of the effectiveness of diagnostic and screening tests. However, Poynard et al. ,137 with reference to biochemical markers of liver fibrosis, have discussed the need for standardisation when comparing the AUROC estimates to take account of differences in the prevalence of stage of fibrosis between studies. 137

Quality of reviews studied

Many of the reviews studied in the preparation of this section referred to an extensive body of research. However, as described above, methodological issues associated with primary studies limit the strength of the conclusions that they can draw.

Furthermore, with respect to identifying non-invasive means of distinguishing NAFLD from NASH, although a large number of techniques have been studied, in general the volume of evidence underpinning individual approaches among individuals with NAFLD appears limited. There is, therefore, a need for additional high-quality research in this area, focusing on those approaches assessed as holding greatest promise.

As described, many of the reviews studied in the preparation of this section referred to a large number of papers. However, the majority contained no description of how these papers were identified, inclusion and exclusion criteria or appraisal methods. In the majority of reviews, it was not clear if the conclusions were based upon an assessment of the quality of the primary studies referred to. Therefore, there is also a need for rigorous systematic reviews of methodologically robust primary research.

Defining ‘non-alcoholic’

Vuppalanchi and Chalasani151 have pointed out the lack of a consistent definition of the level of alcohol consumption compatible with a definition of NAFLD.

Progression to diabetes

The trials have been too short to assess whether or not treatment with any drugs will prevent progression to diabetes.

Which drug?

If there is to be a large trial, would insulin sensitisers be the best drug to test? Rosiglitazone would not be included, because its cardiovascular risk profile makes pioglitazone the drug of choice in the TZDs, and rosiglitazone has been withdrawn from use in the UK. However, pioglitazone is not without risk – oedema, heart failure, fractures. The safety of the glitazones was reviewed in a recent report published in the Health Technology Assessment journal. 84

Metformin, which has been used for more than 50 years, has been proven to be safe in those who can tolerate it – a small proportion have to stop because of gastrointestinal symptoms.

Evidence from the LEAD trials is that liraglutide is effective in improving NAFLD. Armstrong et al. ,141 so far, by abstract only, have reported that many patients in the LEAD trials had advanced NAFLD, but that liraglutide 1.8 mg reduced the NFS score. However, there were no data on the 1.2-mg dose in this abstract. The National Institute for Health and Clinical Excellence (NICE) in a recent appraisal concluded that the 1.8 mg was not recommended by NICE because the marginal benefits did not justify the large increase in marginal cost. 152 In passing, it should be noted that most of these patients were also on metformin, which casts doubt on its long-term effectiveness.

Two recent abstracts provide further information on the effects of GLP-1 analogues on NAFLD. Gardner et al. 153 in a small study gave an unspecified incretin mimetic to seven diabetic patients. Liver fat was reduced from mean 29% to 16%.

Horton et al. 154 pooled data from two DURATION (Diabetes therapy Utilization: Researching changes in A_1c, weight and other factors Through Intervention with Exenatide ONce weekly) trials on 223 patients on once-weekly exenatide and reported improvements in ALT and AST, although values are not given in the abstract.

Pioglitazone

Metformin

For the introduction

• We checked NICE, HTA, SIGN and NHS evidence.

• Searched MEDLINE for general reviews.

• Aetiology, pathology, natural history – searched MEDLINE, limits clinical queries.

• Checked guidelines from the following professional bodies:

  1. – American Gastroenterological Association

  2. – American Association for the Study of the Liver; The British Association for the Study of the Liver

  3. – The British Society of Gastroenterology

  4. – The European Association for the Study of the Liver.

Epidemiology (prevalence and incidence)

• MEDLINE 1990 – current and EMBASE 1990 – current.

• Non-alcoholic fatty liver disease terms plus (incidence or prevalence).tw.

• Web of Science for abstracts.

• Google search for grey literature.

Ovid MEDLINE(R) 1996 to week 3 August 2009

1. exp *fatty liver/

2. (non-alcoholic or non alcoholic or nonalcoholic).tw.

3. (fatty and liver).tw.

4. 3 and 2

5. steato*.tw.

6. 2 and 5

7. (NAFL or NAFLD or NASH).tw.

8. 6 or 4 or 1 or 7

9. limit 8 to (English language and yr=“1999 -Current” and “diagnosis (specificity)”).

Ovid MEDLINE(R) 1950 to week 2 June 2010

1. fatty liver/

2. liver.tw.

3. (non-alcoholic or non alcoholic or nonalcoholic).tw.

4. (fatty or steato*).tw.

5. (NAFLD or NAFLD or NASH).tw.

6. 4 and 3 and 2

7. 6 or 1 or 5

8. exp Metformin/

9. exp Thiazolidinediones/

10. (insulin sensit* or metformin or pioglitazone or rosiglitazone or thiazolidinedione* or glitazone*).tw.

11. 8 or 9 or 10

12. 7 and 11.

Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations, 22 June 2010

1. liver.tw.

2. (non-alcoholic or non alcoholic or nonalcoholic).tw.

3. (fatty or steato*).tw.

4. (NAFLD or NAFLD or NASH).tw.

5. 1 and 2 and 3

6. 4 or 5

7. (insulin sensit* or metformin or pioglitazone or rosiglitazone or thiazolidinedione* or glitazone*).tw.

8. 6 and 7.

EMBASE 1980 to 2010 week 24

1. *fatty liver/

2. liver.tw.

3. (non-alcoholic or non alcoholic or nonalcoholic).tw.

4. (fatty or steato*).tw.

5. (NAFLD or NAFLD or NASH).tw.

6. 2 and 3 and 4

7. 1 or 5 or 6

8. exp metformin/

9. exp thiazolidinediones/

10. (insulin sensit* or metformin or pioglitazone or rosiglitazone or thiazolidinedione* or glitazone*).tw.

11. 8 or 9 or 10

12. 7 and 11.

Science Citation Index – Expanded plus Conference Proceedings Citation Index – Science, 1970 to June 2010

The following conferences were also covered:

• American Gastroenterological Association (Gastroenterology)

• American Association for the Study of Liver Diseases (Hepatology)

• European Association for the Study of the Liver (J Hepatol)

• American College of Gastroenterology (Am J Gastroenterol).

1. TS=liver

2. TS=(non-alcoholic or non alcoholic or ‘nonalcoholic’)

3. TS=(fatty or steato*)

4. #3 and #2 and #1

5. TS=(NAFL or NAFLD or NASH)

6. #5 OR #4

7. TS= (insulin sensit* or metformin or piolitazone or rosiglitazone or thiazolidinedione*)

8. #7 AND #6

9. #7 AND #6 Refined by: Document Type=(MEETING ABSTRACT).

Cochrane

Search alerts

Search alerts were set up in MEDLINE and EMBASE to run daily.

Ovid MEDLINE(R) 1950 to June 2010

1. liver.tw.

2. (non-alcoholic or non alcoholic or nonalcoholic).tw.

3. (fatty or steato*).tw.

4. (NAFL or NAFLD or NASH).tw.

5. exp *Diagnosis/

6. diagnos*.tw.

7. Fatty Liver/di [Diagnosis]

8. 1 and 3 and 2

9. 8 or 4

10. 6 or 5

11. 10 and 9

12. 11 or 7

13. limit 12 to English language.

June 2010

Ovid MEDLINE(R) 1950–2010

1. Fatty Liver/

2. liver.tw.

3. (non-alcoholic or non alcoholic or nonalcoholic).tw.

4. (fatty or steato*).tw.

5. (NAFLD or NAFLD or NASH).tw.

6. 2 and 3 and 4

7. 1 or 5 or 6

8. “Costs and Cost Analysis”/

9. “Cost of Illness”/

10. exp Economics/

11. exp Health Status/

12. exp “Quality of Life”/

13. Quality-Adjusted Life Years/

14. (health state* or health status).tw.

15. (qaly$or EQ5D or EQ-5D or euroqol or euro-qol or SF-36 or SF36).tw.

16. (markov or time trade off or standard gamble or hrql or hrqol or diabilit$).tw.

17. (quality adj2 life).tw.

18. (decision adj2 model).tw.

19. (pharmacoeconomics$or pharmaco-economic$or economics$or cost-effective* or cost benefit).tw.

20. (utilit* adj3 (cost* or analys* or score* or health or value* or assessment*)).tw.

21. 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20

22. 7 and 21.

Web of Science 2010 (meeting abstracts)

TS=liver

TS=(non-alcoholic or nonalcoholic or ‘non alcoholic’)

TS=(fatty or steato*)

#3 AND #2 AND #1

TS=(NAFL or NAFLD or NASH)

#5 OR #4

TS=quality of life

TS=(pharmacoeconomic* or pharmaco-economic* or economic* or cost-effective* or cost–benefit or decision model or health state* or health status)

TS= (qaly$or EQ5D or EQ-5D or euroqol or euro-qol or SF-36 or SF36 or decision model or utilit*)

#9 OR #8 OR #7

#10 AND #6

#10 AND #6 Refined by: Document Type=(MEETING ABSTRACT).