Maddrey, Lille, and Alcohol

Motivation: Compared to hospitals in Baltimore, hospitals in Boston - in my brief experience- appear to get fewer patients with poly-drug abuse but just about the same number of patients with alcohol abuse.  In particular, over the past few months, I have met many patients with alcoholic hepatitis.  Some have died while others have walked out of the hospital against medical advice to the nearest bar.  To distinguish who is likely to get very ill, the Maddrey's Discriminant Function is generally used with severe alcoholic hepatitis defined as a score more than 32 or presence of encephalopathy.  Patients with severe alcoholic hepatitis have significant mortality benefit from early steroid treatment.  Recently, a French group developed a scoring system called the Lille Model, which seeks to better identify patients with poor prognosis even after steroid treatment.  How good is the French system?

Paper: Louvet, A. et. al. The Lille model: a new tool for therapeutic strategy in patients with severe alcoholic hepatitis treated with steroids. Hepatology (2007) 45: 1348-54.

Methods: To identify prognosis in patients with severe alcoholic hepatitis, the study was carried out in two stages.  In the "exploratory" stage, patients with severe alcoholic hepatitis were treated with prednisone 40 mg or IV 32 mg methylprednisolone for 28 days and followed for 6 months.  Severe alcoholic hepatitis was defined as discriminant function [4.6*(pt's PT-control PT) + Tbili] >=32 or presence of encephalopathy.  The derived model was verified on another "validation" cohort with severe alcoholic hepatitis treated with steroids.

Results:
Survival: In total, 320 patients with severe alcoholic hepatitis were included in the "exploratory" cohort.  Survival at one, two, and six months were 86%, 77%, and 65%.  The mean Maddrey Discriminant Function score was 47.5.

Relevant Parameters: In univariate analysis, parameters which are predictive of worse survival outcome after six months are: 1) Age (lower is better), 2) albumin (higher is better), 3) renal insufficiency, 4) difference in bilirubin levels between day 0 and day 7 of treatment, 5) PT time, 6) bilirubin at day 0.  These parameters were then incorporated into a complicated formula called the Lille Model that gives a score between 0 and 1 with higher scores indicating increased probability of death.

Validation and Comparison: 118 patients with severe alcoholic hepatitis requiring corticosteroids were enrolled and followed.  The validity of the model was measured in terms of area under the receiver operating characteristic (AUROC), which is a measure of the performance of the test (basically tests sensitivity versus specificity under varying cutoff scores).  The Lille Model, when compared to other models as indicator of survival at six months, was a better predictor of survival than Discriminant Function or MELD score at presentation.

Number to Remember: The authors next identified a Lille Model score that had the most predictive value for death at six months (optimum balance between sensitivity and specificity).  The optimum value is obtained at Lille Model score of 0.45. Patient with Lille score more than 0.45 had average 6 month survival of 25% compared to average six month survival of 85% for those with scores under 0.45.

Discussion: The Lille Model takes the prediction algorithms one step further.  Currently, the discriminant function just identifies who has severe alcoholic hepatitis and who does not.  With the Lille Model, the likely clinical trajectory of the patient can be more accurately predicted.  Patients with Lille scores above 0.45 have a very high mortality rate that is at times hard to recognize early on!  Hopefully, adjunctive therapies in these patients (like pentoxyfillene) can make a difference.   To me, the most important lesson from the paper is that patients with severe alcoholic hepatitis are potentially very sick.  Even with corticosteroid treatment, about a third will die in six months.  So, recognition of the disease severity and timely treatment are key. 

SIADH and Uric Acid

Motivation: In my brief experience as an intern, most patients predictably have two conditions.  One is a mild anemia (which implies, of course, that you need to have ordered iron, ferritin, folate, and B12 before morning rounds).  The second condition is hyponatremia.  I am not talking about sodium level of 120 but rather about the sodium level of 131.  Since the causes of hyponatremia are so many, there is often no good reflex testing to click.  I was recently told that uric acid is often low in hyponatremia from SIADH.  How sensitive is hypourecemia for hyponatremia?

Paper: Beck, Laurence, "Hypouricemia in the Syndrome of Inappropriate Secretion of Antidiuretic Hormone".  NEJM (1979) 301: 528-30.

Methods: A single person review of records of patients examined by the author (in the grand old tradition).  Over a period of 18 months, patients with hyponatremia (defined as less than 130 mM) without renal failure (Cr<2.0) and not on diuretics were included in the study.  Patients were labeled as having SIADH if they had serum osmolality less than 270 mOsm with urine osmolality greater than 250 mOsm.  Also excluded were patients who were overtly hypervolemic (judged by edema), hypovolemic, or had signs of glucocorticoid deficiency.

Results:
Subjects: In the 18 month period, 49 patients were evaluated for hyponatremia. 19 patients were excluded for lack of uric acid measurement. Seventeen patients met criteria for SIADH wile the remaining 13 patients had euvolemic hyponatremia without SIADH.

Uric Acid Levels: The mean uric acid level in patients with SIADH was 2.9 mg/dL while the patients in the other group had mean level of 7.7 mg/dL (p<0.001).  Only one patient with SIADH had uric acid level greater than 4 mg/dL.  All patients without SIADH had uric acid concentration greater than 5 mg/dL.  In seven patients with SIADH, when fluid restricted, uric acid levels rose to a mean level of 5.2 mg/dL from previous mean of 2.8 mg/dL.

Clearance of Uric Acid: In three patients, clearance of uric acid was measured by 24 hour urine collection.  During periods of hyponatremia from SIADH, the uric acid clearance was increased compared to period after water restriction indicating that overexcretion rather than underproduction was the cause of hypouricemia.

Discussion: The paper is old with many weaknesses like few subjects and lack of testing for hypothyroidism, but this is one of the original papers showing the utility of uric acid measurement in distinguishing SIADH.  One of the key points is that in both hypovolemic and hypervolemic hyponatremia, uric acid is generally expected to be normal or higher than normal.  This paper showed that in euvolemic hyponatremia, SIADH can generally be distinguished from other causes (the paper does not investigate the "other" etiologies) by measuring uric acid levels.  What was so remarkable was the degree of difference in mean uric acid levels among patients with SIADH (mean of 2.9 mg/dL) and those with other causes (7.7 mg/dL).  These findings have been replicated in some subsequent studies (though none have very large number of subjects).  From now on, I think that uric acid level is a great test to send overnight to investigate euvolemic hyponatremia.  

Beta-Blockers in Acute Heart Failure

Motivation: When a patient is admitted with acute heart failure and laboring to breathe, I often debate whether or not to continue the home metoprolol.  In chronic heart failure, I understand the bit about beta-blockade having protective effects.  But, acutely, beta-blockers decrease heart inotropy resulting in decreased cardiac output and elevated left ventricular filling pressures - effects that are not helpful in acute heart failure.  Like so many things in cardiology, it turns out that there has been a randomized trial with a nice acronym (B-CONVINCED) that examines this issue.

Paper:  "B-CONVINCED: Beta-blocker CONtinuation Vs. INterruption in patients with Congestive heart failue hospitalizED for a decompensation episode." Jondeau, G. et. al. European Heart Journal (2009) 30: 2186-2192.

Methods: Randomized, controlled, open-labelled trial conducted in 36 centres in France.  Eligible patients were older than 18 on chronic beta-blocker therapy hospitalized for acute heart failure (including pulmonary edema) with left ventricular ejection fraction less than 40%.  Patients were excluded if found to have STEMI, bradycardia, or second-third degree heart block.  Patients who were initially judged to need dobutamine therapy were also excluded.  Intervention was continuation of home-dose beta-blocker therapy or discontinuation of beta-blocker for at least 3 days.  Primary endpoint was the general health and dyspnea at 3 days of hospitalization. 

Results:
Subjects: Analysis was performed in 147 patients (69 on BB therapy and 78 without BB).  There were no significant differences between groups including age, etiology (ischemic vs non-ischemic), ejection fraction, prevalence of atrial fibrillation, or home CHF treatment regimen.  The most common cause for acute exacerbation was non-adherence to therapy.  The beta-blockers most commonly used were bisoprolol (beta-1 blocker) 70%, carvedilol 11%, and atenolol 10%.  In the group on beta-blocker therapy, beta-blockers were stopped in four patients (three needed dobutamine and one had bronchospasm).

Hospital Courses: There were no significant differences in clinical features of heart failure (dyspnea, pulmonary rales, lower extremity edema, JVD, hepatomegaly) on the first 8 days of hospitalization (including day 3) between the two groups.  Patients without BB had higher heart rates :)  There were no differences in blood pressure during the course of hospitalization in the two groups.  No differences were observed in BNP levels either.  One death occurred in the BB group while two deaths occurred in the BB stopped group (difference not significant).

Long-term Follow-up: At 3 months, death rate and re-hospitalization rate were similar in both groups.  After 3 months, patients who were continued on beta-blockers were more likely to be receiving beta-blockers (90%) than those who had beta-blockers stopped (76%, p=0.04). 

Discussion: This paper shows that even during acute heart failure exacerbations, beta-blockers can be safely continued without adversely affecting rate of recovery.  On the other side of the fence, one can also say that this paper shows beta-blockers can be discontinued during hospitalization without affecting rate of recovery.  Between these two varying viewpoints, I would favor continuing beta-blockers since beta-blockers decrease myocardial ischemia and have proven anti-arrhythmic effect.  Also, as shown in the paper, once beta-blockers are discontinued upon hospitalization, the risk of beta-blockers not being restarted upon discharge increases.

While the paper demonstrates non-inferiority of continuing beta-blockers during acute heart failure, this trial has some important limitations.  First, the study was underpowered to detect potential benefit of beta-blockers in preventing arrhythmias during heart failure.  That part of the benefit of beta-blockers in acute heart failure remains speculative.  Also, in the study, the beta-blocker dosage was not standardized.  It is unclear if beta-blocker titrated to, for example, heart rate is beneficial in one dosage but detrimental at higher doses.