Few studies have determined the natural history of biopsy-proven fatty liver disease and compared the long-term prognosis of these two major groups of fatty liver disease due to alcoholic and non-alcoholic fatty liver disease.
Our study has several methodological strengths. First, all patients had biopsy proven fatty liver disease and histology was re-evaluated based on validated scores. We believe we have included the vast majority of patients with biopsy-proven fatty liver disease in the whole population of Iceland during this period as at least 95% of the population lives in the catchment area of the National University Hospital. The fact that all the patients underwent a liver biopsy is also a weakness, especially when evaluating disease outcome. Previous studies have shown that NAFLD patients diagnosed with liver biopsy have a worse prognosis compared with patients diagnosed with ultrasonography. Therefore, studying patients recruited from the pathology registry involves a selection bias. It must be stressed that the indications for the liver biopsy was not always clear, although most cases were because of elevated liver function tests and/or hepatomegaly or suspected alcoholic liver disease, and this in turn can give a selection bias. Also the indication for biopsy in NAFLD and AFLD may differ between conditions and between practitioners and might explain some of the differences in disease outcome.
The main limitation of the study was its retrospective design and data was not systematically registrered and was therefore sometimes missing or unavailable. In addition the search for the code M-50080 (fatty liver) is limiting in itself as the more serious steatosis with fibrosis and even cirrhosis might be coded as something else than just fatty liver. Since this was a retrospective study on liver biopsies it is not possible to standardize the size of the needle biopsies. Therefore the size is very variable and the range can be considerable, but should be similar to the standards observed in pathology departments in general. Samples less than 2 mm in diameter would have been excluded from the study, but no such samples came into the study.
The slides used for pathological estimation were the original slides and recuts or restaining of slides was not done except for a few exceptional cases where the original slides were unavailable or when a Masson-Trichrome stain had not been performed originally. Occationally the colours of the slides had faided somewhat. This however we do not anticipate having significant effect on the results, especially since ballooning degeneration increases cell size and this is not greatly affected by fading colours.
Another limitation is the small number of hard endpoints with only four patients in the NAFLD group who developed cirrhosis over the follow-up period. There is also a potential uncertainty in the non-histological diagnosis of cirrhosis which must be taken into consideration when reviewing the results.
The information on ASH should be interpreted with caution as NAFLD activity score has to our knowledge not yet been validated in AFLD, but we chose to use it for comparison as the histolopathological development is similar in the two conditions and there is no difference morphologically between NAFLD and AFLD.
One of the main findings in this study was that patients with fatty liver disease showed a markedly higher risk of developing liver-related death compared to the general population. Although significantly higher in the AFLD group liver-related death in the NAFLD group was 7% of all deaths. These findings are in contrast with liver-related death rate in Iceland  which was a mean of 0.1% in the general population during the study period. As in other studies it is a challenge to classify patients into non-alcoholic and alcoholic group. We tried to minimize the misclassification by regrouping those without a known alcohol etiology if the patients were found to have an alcoholic related diagnosis later as for instance alcohol pancreatitis and alcohol dependence. In addition all patients were linked to the database for the National Centre of Addiction Medicine in Iceland.
Liver-related morbidity and mortality
Our results show that patients with NAFLD had a rather benign course in terms of liver-related morbidity and mortality. Only 7% developed cirrhosis after a mean of 13 years of follow-up which is similiar to what previous studies on the prognosis of NAFLD have shown [6, 7, 14–17]. A higher number of patients in the AFLD group developing cirrhosis (20%) after approximately 12 years of follow-up, is also in agreement with previous studies showing worse prognosis in patients with AFLD than NAFLD [8, 9, 18]. These studies have demonstrated that patients with alcoholic fatty liver disease have worse prognosis of their liver disease than patients with NAFLD [8, 9, 18]. In a study on prognosis and life expectancy in chronic liver disease the five year survival was 38% for the alcoholic group and 68% for the non-alcoholic group but only 87% of the patient underwent liver biopsy whereas the rest was diagnosed clinically or with ultrasound . In another study of 7000 patients discharged with the diagnosis fatty liver the mortality was 5.4 fold amongst AFLD and 2.6 amongst NAFLD . In the current study we found that the overall survival was worse in the AFLD group. Patients in the AFLD group had a higher liver-related mortality, but patients in the NAFLD group died more frequently from cardiovascular disease as already demonstrated in previous studies [6, 7, 14, 16].
In the AFLD group the most common cause of death was liver-related (35%).
Other studies have shown that obesity in both NAFLD and AFLD predispose to the development of fatty liver and chronic liver disease [19, 20]. In Iceland, the prevalence of liver cirrhosis due to alcohol is very low, only 3.3/100.000 inhabitants which was 4 times lower than in Sweden . In the current study 45% of the patients in the NAFLD group had a BMI ≥30 and somewhat surprisingly there was no significant difference in BMI between the NAFLD group and the AFLD group. The fact that patients with AFLD did not differ with respect to BMI and incidence of metabolic syndrome-related diseases might reflect a mixed AFLD/NAFLD etiology in the alcoholic group. In the Dionysos study, obesity among heavy drinkers, increased the risk for steatosis by twofold . Moreover, no significant differences were evident concerning conditions associated with metabolic syndrome neither at baseline nor at follow-up, but we had expected a higher portion of diseases associated with metabolic syndrome in the NAFLD group. In a study from Denmark, a significantly higher BMI was seen in NAFLD than in AFLD patients. However, this might at least partly reflect the fact that their patients were recruited from an obesity research project  whereas our patients were unselected patients undergoing a liver biopsy.
In the current study women were in the majority in the NAFLD group but the high proportion of women with fatty liver compared to men may reflect a higher disease burden in women. A recent study from the US also found a higher proportion of women in the NAFLD group  which is in line with our results. A significantly higher prevalence of cirrhosis in female AFLD patients was observed in a Danish study and time to cirrhosis was associated with female gender . Population based studies have shown increased risk of women developing alcohol-induced cirrhosis [23–26].
Progression of NAFLD has been found to be slow and seems to depend a great deal on the initial fibrosis stage [8, 14]. Patients with simple fatty liver at baseline seem to have a good prognosis in terms of liver disease. In a Danish study of 109 patients diagnosed with pure non-alcoholic simple steatosis (without inflammation or significant fibrosis) only one of the patients developed cirrhosis . In the current study more severe lobular inflammation was found in the AFLD group compared to the NAFLD group and a significantly higher number of patients in the AFLD group had steatohepatitis compared to the NAFLD group.
In the total study cohort patients with more severe fibrosis at baseline showed a worse overall survival than patients with none or mild fibrosis at baseline. Based on this we were able to show an association in the total study cohort between the stage of fibrosis and the prognosis. However more NAFLD than AFLD patients had mild or no fibrosis at baseline. This is in agreement with results from a recent study showing that advanced fibrosis in the index liver biopsy was the most important predictor of the prognosis in these patients . A recent Danish study showed that the cirrhosis risk was more than twice as high for the patients with steatohepatitis than for those with pure steatosis .
A Swedish cohort study of patients with biopsy-proven NAFLD and elevated liver tests showed that they had a similar survival compared to the Swedish population . Interestingly the risk of death was increased in patients with non-alcoholic steatohepatitis .
A study from Minnesota, identified 420 patients with NAFLD by imaging or liver biopsy found liver-related complications to be the third most common cause of death among NAFLD patients . This is similar to our results showing liver-related to be the third most common cause of death amongst the NAFLD group. However in the AFLD group liver-related death was the leading cause of death, followed by cardiovascular diseases and malignancy which is in accordance to a previous study where hepatobiliary disease was the leading cause of death in the AFLD .
In the current study, no patient in the NAFLD group died of hepatocellular carcinoma and only one patient in the AFLD group. The absence of HCC among the NAFLD group differs from previous cohort studies showing that 3/129 (2%)  and 2/420 (0.5%)  of NAFLD patients developed HCC . A reasonable explanation for this difference in our study could be that our patients had in general mild changes in the liver biopsy at baseline. It is also conceivable that a longer follow-up time would probably lead to patients with HCC.
In agreement with many previous studies a markedly higher proportion of our women in the NAFLD group died of cardiovascular disease compared to women in the AFLD group, 17 patients vs. one patient respectively. A previous study  showed that cardiovascular disease was the leading cause of death in the AFLD group (men and women toghether), which is at odd with our results. In the current study the leading cause of death in the AFLD group was liver related. We can not find a plausible explanation for this difference, although it has been shown that diagnoses on death certificates can underestimate liver-related mortality  which might have been the case in the Danish study .
The most common cause of death in the NAFLD group was from cardiovascular disease, followed by malignancy which is in agreement with findings of other cohort studies (6, 7, 14, 16,). One study found 34% increased risk of cardiovascular mortality among NAFLD patients compared to the general Minnesota population over a 7.6 year follow-up .