Low-grade steatosis and major changes in portal flow as new prognostic factors in steroid-treated alcoholic hepatitis



The aim of this study was to assess the prevalence and prognostic value of major alterations of portal flow in patients with steroid-treated alcoholic hepatitis. Fifty patients with severe, histologically proven alcoholic hepatitis were enrolled. Clinical data, liver test results, and hepatic Doppler ultrasound findings were collected at inclusion and at month 2. Patients were followed for 1 year or until death. Major changes in portal flow were defined as reversed or alternating flow in the portal trunk and/or in intrahepatic portal branches. Changes in portal flow were observed in 24 (48.0%) of 50 and 17 (39.5%) of 43 patients at inclusion and month 2, respectively. Univariate analysis showed that age older than 50 years, steatosis less than 20% on initial liver biopsy, presence of major changes in portal flow, Child-Turcotte-Pugh score higher than 12, factor V level higher than 45%, and hepatofugal splenic blood flow were associated with a lower 1-year survival. Cox regression analysis showed that steatosis < 20% (relative hazard [RH] = 9.3, P = .0009) and major changes in portal flow (RH = 3.1, P = .04), were independently associated with poor survival. In conclusion, major changes in portal flow are frequent in patients with severe alcoholic hepatitis. Altered portal flow and steatosis < 20% are new prognostic factors in steroid-treated alcoholic hepatitis and must be taken into account in patient management. (HEPATOLOGY 2004;40:1370-1378).

Alcoholic hepatitis (AH) is a necrotizing inflammatory condition that is considered as an acute form of alcoholic liver injury. The severity of AH can be measured using Maddrey's discriminant function (DF, > 32).1 In the absence of treatment, approximately 30% to 40% of patients with severe AH die within 6 months; patients who survive the initial episode of AH have a chance of long-term survival if they remain abstinent.2

Most meta-analyses conclude that glucocorticoids are beneficial in alcoholic hepatitis, with a 1-month survival rate of 65% and 85% in untreated and steroid-treated patients, respectively, suggesting that steroids reduce mortality by approximately 20% to 25%.3 However, the reasons why some patients are nonresponders to steroid therapy are unclear. Factors other than necrotizing inflammatory factors could be responsible for the deterioration of liver function despite both abstinence and steroid therapy. Recently, we observed that reversal of portal flow through large spontaneous portacaval shunts in candidates for liver transplantation with alcoholic cirrhosis was more frequent among patients presenting with severe liver insufficiency than in those presenting with complications of portal hypertension.4 We therefore postulated that such major changes in portal flow might contribute to the deterioration of liver function in patients with alcoholic hepatitis. To test this hypothesis, we conducted a multicenter prospective study designed to assess the prevalence and characteristics of major changes in portal flow in patients with steroid-treated severe AH, and to identify variables associated with survival.


AH, alcoholic hepatitis; DF, discriminant function; PT, prothrombin time; CTP, Child-Turcotte-Pugh; US, ultrasound; PV, main portal vein; RPV, right intrahepatic portal vein; LPV, left intrahepatic portal vein; SV, splenic vein; SMV, superior mesenteric vein.

Patients and Methods

The study involved patients hospitalized for severe AH between November 1999 and June 2000 in the liver units of 5 hospitals located in the southeastern suburbs of Paris, France (Hôpital Henri Mondor, Centre Hospitalier Intercommunal, and Hôpital Albert Chenevier in Créteil, Hôpital Jean Verdier in Bondy, and Centre Hospitalier de l'Agglomération Montargoise in Montargis). The study was approved by the Institutional Review Board before its initiation, and written informed consent was obtained from all participating patients.

Inclusion criteria were alcohol intake greater than 40 g/d for at least 1 year before admission, active drinking before admission, histologically proven acute AH, Maddrey DF higher than 32, and/or hepatic encephalopathy. Enrollment in the study was attempted within the first 5 days after admission and was always preceded by transvenous liver biopsy. Exclusion criteria were concomitant bacterial infections, active gastrointestinal hemorrhage, severe cardiovascular or pulmonary disease, renal failure (serum creatinine > 130 μmol/L), sonographic signs consistent with hepatocellular carcinoma, portal vein thrombosis, a history of surgical portacaval shunting, and additional causes of liver disease, namely chronic hepatitis B or C infection, genetic hemochromatosis, and consumption of hepatotoxic drugs. All patients were screened by one of the investigators and by a senior staff member of each liver unit before being enrolled in the study.

From November 1998 to June 2000, 64 patients with a diagnosis of severe AH were considered for enrollment. Fifty patients complying with the inclusion and exclusion criteria were finally enrolled.

Sample Size Calculations.

The number of patients had been calculated under the following hypotheses:

  • Prevalence of major alterations in portal flow: 50%. This figure was chosen based on the results of a previous systematic study of portal flow in 39 liver transplant candidates for alcoholic cirrhosis with end-stage liver diseases,4 in which we observed a 56.5% prevalence of major changes in portal flow in a subgroup of patients listed with a pattern of severe liver insufficiency;

  • Expected 1-year mortality rates: 50%, if major changes in portal flow were present, and 20% if such changes were absent;

  • Alpha and beta risks: 5% and 20% respectively, 1-tailed test.

Under these hypotheses, 62 patients were found to be necessary. At the term of an 18-month, period of inclusion, 50 patients had been enrolled and it was decided to stop the inclusions at this figure.

Prednisolone, 40 mg/d for 1 month, was initiated once the diagnosis of severe acute AH was established. At inclusion (M0) and after 2 months of follow-up (M2), age, sex, alcohol consumption (g/d), duration of steroid therapy, and, when appropriate, the cause of steroid withdrawal (death, noncompliance, infection, etc.), abstinence after admission, treatment with beta-blockers, presence of esophageal varices (graded in 3 stages),5 and signs of ascites and hepatic encephalopathy on physical examination were assessed.

Laboratory tests, including a blood cell count, liver biochemistry, electrolytes, blood urea nitrogen, serum creatinine, fasting blood glucose, albumin, the prothrombin time (PT), and factor V activity, were performed at M0 and M2. The length of follow-up was recorded until death or last news, and 1-year (M12) survival was assessed. Liver disease severity was scored using the Child-Tucotte-Pugh (CTP) classification.6 The severity of AH was defined by the presence of a DF >32 and/or hepatic encephalopathy.

Liver biopsy specimens were interpreted by a single pathologist in each center. The necroinflammatory index was classified in 3 stages (minimal, moderate, and severe), and fibrosis was classified in 4 grades (minimal F1, moderate F2, severe F3, and cirrhosis F4) by extrapolation from the METAVIR scoring system.7 Steatosis was staged in 2 classes (absent or mild [<20%] vs. moderate or severe [>20%]). Polymorphonuclear cell infiltration (minimal or moderate vs. marked) and Mallory bodies (absent or rare vs. numerous) were also evaluated semiquantitatively in 2 classes. Alcoholic hepatitis was defined histologically by the combination of polymorphonuclear cell infiltration and hepatocyte injury. Mallory bodies and steatosis were not necessary for the diagnosis of AH.

Characteristics of the Study Population.

Demographic, clinical, biochemical, endoscopic and histological findings at enrollment are shown in Table 1. Overall, liver function was severely impaired, as shown by a mean PT of 36.3% ± 9.7%, a mean total serum bilirubin value of 199.3 μmol/L ± 130.1 μmol/L, a mean Maddrey's DF of 53.4 ± 19.1, and a mean CTP score of 11.4 ± 1.4.

Table 1. Characteristics of the 50 Patients at Enrollment
  1. NOTE. Values are mean ± SD.

  2. Abbreviations: PMN, polymorphonuclear cell; AST, aspartate aminotransferase; ALT, alanine aminotransferase; N, normal value; GGT, gamma-glutamyltransferase.

Age (yr)52.0 ± 9.6
Male/female, n (%)29 (58)/21 (42)
Alcohol intake (g/d)115.4 ± 57.3
Body mass index (kg/m2)22.9
Maddrey's DF53.4 ± 19.1
CTP score11.4 ± 1.4
Esophageal varices, absent-small/medium-large (%)23 (46)/27 (54)
Ascites/encephalopathy, n (%)37 (74)/10 (20)
Leukocytes/PMN (/μL)10,486 ± 4,753/7,693 ± 4,222
Hemoglobin (g/dL)10.9 ± 2.1
Platelet count (/μL)122,000 ± 57,900
AST/ALT (IU/L) [N < 40/45]145.8 ± 218.5/67.4 ± 150.7
Bilirubin, total/conjugated (μmol/L)199.3 ± 130.1/126.0 ± 107.2
GGT/alkaline phosphatases (IU/L) [N < 65/115]294.5 ± 332.8/155.4 ± 99.9
Prothrombin time/factor V (%)36.3 ± 9.7/46.0 ± 17.2
Albumin (g/L)27.9 ± 5.8
Creatinine (μmol/L)71.4 ± 30.6
Histological Data, n (%) 
 Necroinflammatory activity, mild-moderate/severe33 (66)/17 (34)
 Fibrosis, F3/F46 (12)/44 (88)
 Steatosis, <20%/>20%16 (32)/34 (68)
 Mallory bodies, absent-rare/numerous27 (54)/22 (46)
 PNM infiltrate, mild-moderate/important28 (56)/22 (44)
Major changes in portal flow, n (%)24 (48.0)
Portal vein diameter (mm)10.2 ± 1.8
Portal vein velocity (cm/s)6.6 ± 7.5

Study of Portal Flow.

Doppler examination of the liver was performed in all cases at inclusion (M0) and at M2 in patients who were alive at this time point. The same device was used in the 5 centers (128XP or Sequoia 512, ACUSON, Mountain View, CA), with a multifrequency probe. Doppler ultrasound (US) was performed in each center by a single radiologist blinded to the patients' clinical and biochemical results.

Doppler US was done in the fasting state, in dorsal decubitus, 24 hours after large-volume paracentesis when required. A morphological study was done first. Color Doppler was used to study the direction of flow in the main portal vein (PV), the right intrahepatic portal vein (RPV) and left intrahepatic portal vein (LPV) branches, the splenic vein (SV), and the superior mesenteric vein (SMV). For each vessel, the patient's position and the probe angle were adjusted to obtain a suitable probe-vessel angle and an acceptable color signal. The patient was asked to breathe softly, so that the direction of flow in each vessel could be determined from the color signal (hepatopetal, hepatofugal, or to-and-fro). To-and-fro flow was recorded when the flow was hepatopetal during inspiration and hepatofugal during expiration. We also recorded the presence of direct and indirect portosystemic shunts. For statistical analysis, to-and-fro flow in the Pv, and the combination of hepatopetal portal flow, reversed flow in the RPV, and repermeabilization of the paraumbilical vein were assimilated to hepatofugal flow in the portal trunk.

For quantitative analysis of portal flow, several pulsed Doppler recordings were made in the midsection of the main portal vein. The patient's position and the probe angle were adjusted to obtain the smallest possible probe-vessel angle (always less than 50°). The recording was made in breath-hold conditions, with a gate adapted to the size of the vessel and appropriate recording parameters. Mean flow velocity was calculated by averaging 3 measures of maximal velocities calculated from envelope of the Doppler trace, taking into account the probe-vessel angle. Three measures were accepted if variability was less than 10%. Mean velocity was arbitrarily set at 0 cm/s for reversed and to-and-fro flow.

Statistical Analysis.

The baseline characteristics of the patients, and other continuous variables, are expressed as mean ± 1 SD. The distribution of categorical variables was compared by using the chi-square test or Fisher exact test as appropriate. Quantitative variables were compared using Student t test or the nonparametric Wilcoxon test as appropriate. P values less than .05 were considered significant. The probabilities of survival were calculated by the Kaplan-Meier method and compared by the mean of the log-rank test. Univariate analysis was used to identify significant differences between survivors and nonsurvivors and to compare 1-year probabilities of survival. Stepwise logistic regression and Cox proportional hazards regression analysis were then used to identify factors associated independently with survival (BioMedical Data Package statistical software, Berkeley, CA); variables related to survival in univariate analysis (P < .05) were included in the models.


Prevalence and Characteristics of Altered Portal Flow at Enrollment.

Blood flow could be studied in the PV and LPV of all the patients, and in the RPV of all but 1 of the patients (96%). PV flow was hepatopetal in 28 cases (56.0%), hepatofugal in 19 cases (38.0%), and alternated in 3 cases (6.0%). Periumbilical vein patency was studied in 48 of 50 cases. Blood flow was observed in the periumbilical vein in 15 (31.2%) of 48 cases. Reversed (n = 18) or alternating (n = 7) flow in the RVP was observed in 25 (50.0%) of 50 cases. In 2 of these cases, flow reversal in the RVP coexisted with hepatopetal flow in the PV, due to periumbilical vein patency. This association was considered equivalent to reversal of portal flow in the main PV, therefore signifying a major change in portal flow.

Reversed (n = 16) or alternating (n = 7) flow in the LVP was observed in 23 (46.0%) of 50 cases. Overall, a major change in portal flow was observed in 24 patients (48.0%) (hepatofugal in 19, alternating in 3, hepatopetal portal flow but hepatofugal intrahepatic right portal flow with periumbilical vein patency in 2 cases).

Blood flow could be studied in the SV and SMV in 46 cases (92.0%) and 35 cases (70.0%), respectively. Total reversal of flow in the SV and SMV was observed in respectively 12 (26.1%) and 7 (20.0%) cases, respectively. The reversal of splenic blood flow was observed in 8 (42.1%) of 19 patients with reversed portal flow and in 4 (14.8%) of 27 patients with hepatopedal flow, P = .05. The reversal of mesenteric blood flow was observed in 6 (50%) of 12 patients with reversed portal flow and in 1 (4.3%) of 23 patients with hepatopedal portal flow, P = .003.

Changes in Portal Flow Characteristics Between M0 and M2, and Relationship With Liver Function.

Six patients died between M0 and M2, a median of 21 days after enrollment (range, 2-50 days), and 1 patient was lost to follow-up. Among these 6 patients who had died, portal blood flow on admission was hepatofugal in 4 and hepatopetal in 2. The remaining 43 patients had the prospective Doppler US evaluation at M2. Twenty-one of these patients had major changes in portal flow at M0.

Among the 21 patients with major changes in portal flow at M0, hepatofugal flow became hepatopetal at M2 in 4 cases (19.0%), and alternating portal flow became hepatopetal in 2 cases (9.5%), resulting in the disappearance of major alterations in 28.5% of the 21 patients concerned. In 2 additional cases (9.5%), hepatofugal flow at M0 switched to alternating at M2, but this was considered as a persistent severe alteration of portal flow. The improvement in extrahepatic portal flow was accompanied by an improvement in intrahepatic portal flow in all but 1 of the patients in whom portal flow improved from hepatofugal to alternating. At M2, liver function tended to be better in patients with improved portal flow compared to patients with no improvement, as reflected by the serum bilirubin level (31.8 μmol/L ± 13.0 μmol/L vs. 88.3 μmol/L ± 68.2 μmol/L, P = .07), the prothrombin time (56.7% ± 12.2% vs. 41.4% ± 15.9%, P = 0.1) and Maddrey's DF (20.1 ± 10.4 vs. 45.4 ± 29.5, P = 0.06). Similarly, among the patients with ascites on M0, disappearance of ascites was observed more frequently on M2 in patients with an improvement in portal flow than in patients without such an improvement (5 of 6 vs. 1 of 6, P = .015, Fisher exact test).

Among the patients with major changes on M0, the improvement of portal flow was observed to the same extend at M2, whether the patients had returned to drinking or not (data not shown). Among the 22 patients with no major changes in portal flow at M0, portal flow had deteriorated at M2 in 2 cases. In 1 case, initial hepatopetal flow became alternating; in the other case, intrahepatic right portal flow became hepatofugal. Overall, major changes were observed in 17 (39.5%) of 43 patients at M2.

Baseline Variables Associated With 1-Year Survival.

At 1 year, 4 patients (8%) had been lost to follow-up; information on causes and dates of death were obtained from the registries of the patients' birth towns. Overall (Fig. 1), 34 patients (68%) were alive and 16 patients (32%) had died. Causes of death were end-stage liver failure in 11 cases, hepatorenal syndrome in 2 cases, sepsis in 1 case, variceal bleeding in 1 case, and other cause not related to liver in 1 case.

Figure 1.

Overall 1-year survival among 50 patients with severe alcoholic hepatitis treated with corticosteroids.

Univariate Analysis.

Survivors and nonsurvivors at M12 are compared in Table 2. On admission, nonsurvivors were significantly older and had significantly lower PT values and portal velocity than survivors. Among nonsurvivors, the proportions of patients with major changes in portal flow (68.7% vs. 38.2%, P = .044), with hepatofugal splenic flow (53.8% vs 15.1%, P = .02) and low-grade steatosis (i.e.,<20%) at pretreatment liver biopsy (68.7% vs. 14.7%, P = .001) were significantly higher than among survivors. In addition, the composite Maddrey DF was significantly higher in nonsurvivors than in survivors. The peripheral blood neutrophil count and the CTP score tended to be higher in nonsurvivors than in survivors.

Table 2. Baseline Factors Associated With 1-Year Survival: Univariate Analysis
 Alive at M12 (n = 34)Dead at M12 (n = 16)P
  1. NOTE. Values are mean ± SD.

  2. Abbreviations: PMN, polymorphonuclear cell; AST, aspartate aminotransferase; N, normal value; ALT, alanine aminotransferase; GGT, gamma-glutamyltransferase.

Age (yr)49.9 ± 9.656.4 ± 8.1.024
Male, n (%)19 (55.9)10 (62.5).65
Alcohol intake (g/d)119 ± 59108 ± 54.54
Maddrey's DF49.1 ± 14.762.6 ± 24.1.017
CTP score11.1 ± 1.511.9 ± 1.1.06
Esophageal varices, medium-large (%)18 (52.9)9 (56.3).82
Ascites (%)26 (76.5)11 (68.8).56
Encephalopathy (%)7 (20.6)3 (18.8).87
Leukocytes (/μL)9,847 ± 4,68611,843 ± 4,752.16
PMN (/μL)6,944 ± 3,9419,237 ± 4,177.07
Hemoglobin (g/dL)11.1 ± 2.210.8 ± 2.0.67
Platelet count (/μL)123,000 ± 59,000120,000 ± 57,000.84
AST (IU/L) [N < 40]159 ± 264117 ± 48.53
ALT (IU/L) [N < 45]72 ± 18157 ± 39.74
Total serum bilirubin (μmol/L)200 ± 130198 ± 135.95
GGT (IU/L) [N < 65]346 ± 337239 ± 326.26
Alkaline phosphatases (IU/L) [N < 115]159 ± 117147 ± 46.67
Prothrombin time (%)38.4 ± 9.431.9 ± 9.0.026
Factor V (%)48.5 ± 17.340.7 ± 16.3.14
Albumin (g/L)28.7 ± 6.626.1 ± 2.9.13
Creatinine (μmol/L)67.5 ± 21.478.9 ± 44.2.18
Necroinflammatory activity, mild-moderate (%)22 (64.7)11 (68.7).77
Fibrosis score F4, n (%)29 (85.3)15 (93.8).40
Steatosis, absent or mild, n (%)5 (14.7)11 (68.8).001
Mallory bodies, numerous, n (%)14 (41.2)9 (56.3).31
PMN infiltrate, mild-moderate, n (%)21 (61.8)7 (43.8).23
Major changes in portal flow, n (%)13 (38.2)11 (68.8).044
Hepatofugal splenic flow, n (%)7/13 (53.8)5/33 (15.1).02
Hepatofugal mesenteric flow, n (%)3/8 (37.5)4/27 (14.8).31
Portal vein diameter (mm)10.4 ± 1.89.6 ± 1.8.16
Portal vein velocity (cm/s)8.3 ± 7.92.8 ± 4.9.022

Using survival analysis techniques, 6 baseline variables were associated with statistically different 1-year survivals (log-rank test): age > 50 or < 50 (56.7% vs. 85.0%, P =.034); steatosis < 20% or > 20% (31.3% vs. 85.2%, P < .0001); presence of major changes in portal flow (54.2% vs. 80.8%, P = .05); CTP score > 12 or < 12 (53.8% vs. 83.3%, P = .02); and factor V level < 45% or > 45% (54.2% vs. 80.8%, P = .034). The 1-year survival was also significantly lower in patients with hepatofugal splenic flow compared to patients with hepatopedal splenic flow (41.7% vs. 82.4%, P = .004) but did not differ between patients with hepatofugal or hepatopedal flow in the mesenteric vein (57.1% vs. 82.1%, P = .13).

Among the 44 patients surviving at M2, 28 patients had remained abstinent and 16 patients had returned to drinking. The proportion of patients returning to drinking was similar in M12 survivors or nonsurvivors (11 [32.3%] of 34 vs. 5 [50.0%] of 10, P = .45).

Multivariate Analysis.

Stepwise logistic regression analysis of variables that differed between survivors and nonsurvivors (P < .05) in univariate analysis showed that three variables had independent prognostic value, namely steatosis less than 20% at pretreatment liver biopsy (OR, 21.1; 95% CI, 3.2-137.2; P = .0001), major changes in portal flow (OR, 6.6; 95% CI, 1.1-41.3; P = .021), and age over 50 years (OR, 5.4; 95% CI, 0.9-34.30; P = .047). Using Cox proportional hazards regression, low-grade steatosis (relative hazard [RH], 9.3; 95% CI, 2.48-34.61; P = .0009), and presence of major changes in portal flow (RH: 3.1; 95% CI, 1.04-9.18; P = .04) had independent prognostic value (Fig. 2). Age did not enter the model (P = .078).

Figure 2.

Impact of the 2 independent variables influencing 1-year survival (log-rank test). (A) Influence of major changes in portal blood flow. Black line indicates no major changes (n = 26) and gray line indicates presence of major changes (n = 24). (B) Influence of steatosis on pretreatment liver biopsy. Black line indicates moderate to marked steatosis (>20%, n = 34) and gray line indicates no or mild steatosis (<20%, n = 16).

Comparison of Patients With and Without Major Changes in Portal Flow (Table 3).

Patients with major changes in portal flow had significantly higher neutrophil counts, a lower main PV diameter, and lower portal velocity than patients without such changes. The 2 groups were similar with regard to all other variables.

Table 3. Patient Comparison According to Major Changes in Portal Flow
 Patients With Major Changes (n = 24)Patients Without Major Changes (n = 26)P
  1. NOTE. Values are mean ± SD.

  2. Abbreviations: PMN, polymorphonuclear cell; AST, aspartate aminotransferase; N, normal value; ALT, alanine aminotransferase; GGT, gamma-glutamyltransferase.

Age (yr)52.9 ± 8.251.1 ± 1.8.51
Male/female, n (%)14 (58)/10 (42)15 (58)/11 (42)>.9
Alcohol intake (g/d)115 ± 57116 ± 58.92
Maddrey's DF54.2 ± 19.952.7 ± 18.6.78
CTP score11.5 ± 1.311.3 ± 1.5.50
Esophageal varices, absent-small/medium-large (%)12 (50)/12 (50)11 (43)/15 (57).80
Ascites, n (%)15 (62.5)22 (84.6).14
Encephalopathy, n (%)6 (25.0)4 (15.4).6
Leukocytes (/μL)11,608 ± 4,7739,450 ± 4,581.10
PMN (/μL)8,903 ± 4,2316,530 ± 3,734.043
Hemoglobin (g/dL)11.1 ± 2.010.9 ± 2.2.06
Platelet count (/μL)135,400 ± 62,600110,000 ± 51,400.12
AST (IU/L) [N < 40]114 ± 43179 ± 302.33
ALT (IU/L) [N < 45]53 ± 3380 ± 207.53
Total serum bilirubin (μmol/L)196 ± 119202 ± 142.87
GGT (IU/L) [N < 65]239 ± 326346 ± 337.26
Alkaline phosphatases (IU/L) [N < 115]152 ± 77158 ± 117.83
Prothrombin time (%)34.9 ± 9.037.6 ± 10.3.34
Factor V (%)43.1 ± 16.348.7 ± 17.9.25
Albumin (g/L)27.4 ± 5.528.3 ± 6.1.6
Creatinine (μmol/L)67.8 ± 23.474.8 ± 36.2.42
Portal vein diameter (mm)9.3 ± 1.510.9 ± 1.8.003
Necroinflammatory activity, mild-moderate/severe (%)15 (62)/9 (38)18 (69)/8 (31).83
Steatosis, <20%/>20% (%)8 (33)/16 (67)8 (31)/18 (69)>.99
PNM infiltrate, mild-moderate/important (%)12 (50)/12 (50)16 (61)/10 (39).60
Mallory bodies, absent-rare/numerous (%)13 (54)/11 (46)14 (54)/12 (46)>.99

Comparison of Patients With Steatosis Less Than 20% or Greater Than 20% (Table 4).

Compared to patients with moderate to severe steatosis, patients with low-grade steatosis tended to have lower alcohol intake before admission (95.0 g/L ± 33.9 g/L vs. 125.0 g/L ± 63.7 g/L, P = .08), although this difference did not reach statistical significance. They also had significantly more severe liver function impairment, with a lower PT, a lower factor V level (31.8% ± 8.8% vs. 38.5% ± 9.5%, P = .02, and 36.9% ± 12.1% vs. 50.3% ± 17.7%, P = .009, respectively), and a higher Maddrey DF (61.4 ± 22.5 vs, 49.7 ± 16.3, P = .04). All other parameters were similar in the 2 groups.

Table 4. Features of the Patients According to the Presence of Pretreatment Steatosis
 No or Mild Steatosis (<20%) (n = 16)Moderate to Marked Steatosis (n = 34)P
  1. NOTE. Values are mean ± SD.

  2. Abbreviations: PMN, polymorphonuclear cell; AST, aspartate aminotransferase; N, normal value; ALT, alanine aminotransferase; GGT, gamma-glutamyltransferase.

Age (yr)53.6 ± 9.151.2 ± 9.8.40
Male/female, n (%)10/619/15.65
Alcohol intake (g/d)95.0 ± 33.9125.0 ± 63.7.084
Body mass index (kg/m2)24.3 ± 3.522.4 ± 3.4.12
Diabetes mellitus, n (%)2 (12.5)3 (8.8).5
Maddrey's DF61.4 ± 22.549.7 ± 16.3.04
CTP score11.4 ± 1.711.4 ± 1.3.93
Esophageal Varices, absent-small/medium-large (%)5 (31)/11 (69)18 (53)/16 (47).15
Ascites, n (%)10 (62.5)27 (79.4).20
Encephalopathy, n (%)3 (18.7)7 (20.6).87
Leukocytes (/μL)11,194 ± 5,34110,152 ± 4,497.47
PMN (/μL)8,548 ± 5,0557,278 ± 3,600.31
Hemoglobin (g/dL)10.8 ± 2.111.1 ± 2.1.7
Platelet count (/μL)104,000 ± 48,000156,000 ± 257,000.12
AST (IU/L) [N < 40]122 ± 82114 ± 43.62
ALT (IU/L) [N < 45]50 ± 3175 ± 182.58
Total serum bilirubin (μmol/L)174.7 ± 119.0210.8 ± 135.2.36
GGT (IU/L) [N < 65]251.4 ± 318.4314.8 ± 342.1.53
Alkaline phosphatases (IU/L) [N < 115]141 ± 42162 ± 117.49
Prothrombin time (%)31.8 ± 8.838.5 ± 9.5.022
Factor V (%)36.9 ± 12.150.3 ± 17.7.009
Albumin (g/L)28.7 ± 5.527.6 ± 5.9.52
Creatinine (μmol/L)83.5 ± 43.365.8 ± 20.9.056
Mild or moderate necroinflammatory activity, n (%)11 (68.7)22 (64.7).77
Fibrosis score F4, n (%)15 (93.7)29 (85.3).39
Numerous Mallory bodies, n (%)8 (50.0)15 (44.1).69
Mild or moderate PMN infiltrate, n (%)8 (50.0)20 (58.9).55
Major changes in portal flow, n (%)8 (50.0)16 (47.0).71


Subject to the small number of events observed in the present work that could affect the precision of analyses, this study yielded 3 new findings in severe AH: (1) major alterations of portal flow are common in patients with severe AH, (2) severe alterations of portal flow have a negative impact on survival among patients with severe AH, and (3) low-grade steatosis is associated with a specific pattern of AH, defined by steatosis less than 20% at pretreatment liver biopsy, severely impaired liver function, and a poor prognosis.

The prevalence and consequences of major changes in portal flow have not been previously studied in patients with severe histologically proven AH. Such changes were observed in 48% of our patients at admission. In previous reports, the prevalence of hepatofugal portal flow in patients with cirrhosis ranged from 3.1% to 22.5%.8–12 This wide range of values probably reflects not only the use of different techniques to assess portal flow, but also differences in the causes and stages of liver disease.11–12 Also, the highest prevalence of portal flow reversal reported so far (i.e., 22.5%)12 was observed in patients with alcoholic cirrhosis and CTP class C. Our study suggests that the prevalence of such abnormalities is more frequent in patients with severe AH. This could be related to an increase in portal pressure due to superimposed AH. However, in a previous study of Doppler US portal hemodynamic changes in liver transplant candidates for alcoholic cirrhosis,4 we observed a 56.5% prevalence of major alterations of portal flow in a subgroup of patients listed with a prominent pattern of severe liver insufficiency. This suggests that the frequency of major alterations of portal blood flow might have been underestimated in the very peculiar settings of major liver insufficiency, as observed in severe AH and in liver transplant candidates.

The second finding in this study was the negative impact of portal flow reversal on survival. Recently, several prognostic factors have been identified in patients treated with steroids for severe AH, including age,13 variceal bleeding, renal failure, and the neutrophil count.14 To study the prognostic impact of portal flow more specifically, we excluded patients with bleeding and renal failure. Our study confirmed the prognostic role of age and identified 2 new variables associated with a poorer prognosis: major alterations of portal flow, and low-grade steatosis at pretreatment liver biopsy.

The prognostic value of portal hemodynamic changes was based on (1) the results of the multivariate analysis showing the independent detrimental prognostic value of major changes in portal flow and (2) the improvement in liver function and clinical condition of the patients in whom an improvement in portal flow had occurred between M0 and M2.

In approximately 30% of the cases in this series, resolution of AH might have contributed to both the restoration of portal blood flow toward normal and the improvement in liver function. However, in the remaining patients (70%) with major changes on admission, portal flow remained severely impaired on M2, probably because of large spontaneous portacaval shunts, and this latter situation was found to be associated with poorer liver function and lower survival.

The prognostic value of portal flow reversal in AH concurs with some clinical and experimental data supporting a role of portal flow reversal in the deterioration of liver function. In patients with cirrhosis, a large spontaneous portacaval shunt, reduces hepatic synthesis and clearance functions.15 Also, after surgical side-to-side portacaval shunting, the direction of portal flow predicts patient survival independently of the severity of liver failure before surgery,16 and liver atrophy can be aggravated by surgical portacaval shunting.17 In rats, hepatic regeneration in response to partial hepatectomy was found to be significantly less efficient in animals with an end-to-side portacaval shunt than in animals with no shunt.18 Preliminary results also suggest that restoring hepatopetal flow by occluding a spontaneous portacaval shunt19 and increasing portal flow by implanting a portal pump20 may improve hepatic synthesis and clearance functions. This is supported by the fact that liver function tended to improve in our patients in whom portal flow improved between M0 and M2. Finally, our results are in agreement with the relationship between portal hemodynamics, as assessed by Doppler US parameters, and the prognosis of cirrhosis, as already reported in the literature.21–25

Our last finding was a specific pattern of severe AH characterized by low-grade (<20%) steatosis on liver biopsy and a very poor prognosis. Indeed, the 1-year survival rate among the 16 patients with this pattern was only 31%, compared with 85% among the 34 patients with moderate or marked steatosis (P < .0001, Fig. 2). Why steatosis should be less pronounced in some patients than in others is unclear, although the extent of recent alcohol abuse or abstinence, more advanced liver disease with gradual disappearance of steatosis (as observed in the natural history of nonalcoholic steatohepatitis26), and a reduced capacity to synthesize triglycerides could be involved. We therefore compared the patients with low-grade or marked steatosis. All patients were active drinkers on admission, and transvenous biopsy was always performed within the 5 first days following admission, ruling out the responsibility of abstinence on the extent of steatosis. Patients with low-grade steatosis tended to have lower, but not statistically different, daily alcohol intake before admission, significantly lower prothrombin and factor V values, and a significantly higher Maddrey DF. These results reflect a more advanced liver failure in low-grade steatosis patients than in other patients and suggest that low-grade steatosis steatosis may be a marker for patients who deteriorate despite abstinence and are potentially more ill. These results also suggest that low-grade steatosis results in lower sensitivity to steroid therapy. With respect to this last point, it has been shown in a model of ischemia-reperfusion injury in steatotic livers that steatotic hepatocytes are more susceptible to the proapoptotic properties of tumor necrosis factor α than are nonsteatotic hepatocytes.27 Thus, in severe AH, the antiapoptotic actions of steroids may be more pronounced in patients with marked steatosis than in patients with little or no steatosis, resulting in poorer efficacy in the latter group.

Four randomized studies have shown a beneficial impact of steroids on the survival of patients with AH,28–31 yet efficacy remains controversial. Discordant results obtained in other studies32–40 could be due to false-positive diagnosis of patients with no pathological assessment before treatment, or to differences in steroid regimens, or to a lack of adjustment for prognostic variables such as renal failure and bleeding. Our results show that other variables, namely portal flow reversal and steatosis, can influence survival and should be taken into account.

In conclusion, this prospective multicenter study shows that severely altered portal flow is common in patients with severe AH. It also identified 2 new variables associated with poor outcome, namely major alterations of extrahepatic or intrahepatic portal flow, and low-grade steatosis on pretreatment liver biopsy. These parameters should therefore be taken into account when assessing steroid efficacy. We recommend routine Doppler US of the liver and transvenous liver biopsy at admission for patients with severe AH.