Wednesday, April 25, 2012

The Demise of Protime Clinic?

The last remaining market for warfarin?
The final results of the EINSTEIN-PE trial were published in the New England Journal of Medicine this month.  EINSTEIN-PE was a randomized, open-label trial pitting rivaroxaban, a factor Xa inhibitor, against enoxaparin overlapping with warfarin (or other vitamin K antagonist) in adults with symptomatic pulmonary embolism.  The primary efficacy endpoint was recurrent venous thromboembolism.  

EINSTEIN-PE was a noninferiority study, which are all the rage these days.  Rivaroxaban would be deemed noninferior if the upper limit of the 95% confidence interval of the hazard ratio was 2 or less.  Put another way, rivaroxaban would be considered as good as warfarin as long as it was statistically unlikely to be more than twice as bad.  (Go ahead and thank us for clearing that up!)

Setting aside the "noninferiority" descriptor for a moment, let's take a look at the results:  1.8% of patients on warfarin had a recurrent VTE, compared to 1.8% of those on rivaroxaban, a non-significant difference.  Of course, with a negative study (i.e., one showing no difference in outcomes), you always want to look at the width of the 95% confidence interval, to make sure the trial wasn't simply underpowered.  The absolute risk reduction was 0.3% in favor of warfarin, with a pretty tight 95% CI of -0.5% to 1%.  (The authors don't report the data in terms of ARRs; I used an online calculator.)  Another way of stating the results:  Warfarin is unlikely to be more than 1% better than rivaroxaban, or, if you're a glass half-empty kind of person, rivaroxaban is unlikely to be more than 1% worse than warfarin in preventing recurrent VTE.  

The other major finding was that rivaroxaban had a lower rate of major bleeding of 1.1%, compared to 2.2% in the standard treatment group.  The investigators' definition of noninferiority is almost irrelevant.  Practically speaking, most doctors would agree that a drug with an almost equivalent rate of recurrent VTE and a decreased risk of major bleeding should be considered at least as good as standard therapy, especially without the added hassle of blood tests and dosage adjustments.

So why aren't more physicians hopping on the rivaroxaban bandwagon?  Don't worry, they will, but it will take time -- specifically, the amount of time it takes for rivaroxaban to go generic, or to drop in price.  We're talking about a medication that costs approximately $8 a day.  It's still not FDA-approved for treatment of VTE*, so it may not be covered by many insurance companies.

And what about rivaroxaban's other major competitor, the direct thrombin inhibitor, dabigatran?  There haven't been any head-to-head studies, and the pharmaceutical companies are smart enough not to fund one.  That said, I predict that rivaroxaban will end up trouncing dabigatran because 1) it doesn't have that pesky dyspepsia side effect and 2) you don't have to overlap it with heparin.  (RE-COVER, the trial comparing dabigatran to warfarin for VTE, used parenteral anticoagulation in both groups to maintain blinding, even though dabigatran has a fast onset of action.  Boehringer Ingelheim is probably kicking itself for being so scientifically rigorous.)

To all of you internists out there -- what's your first-line drug for venous thromboembolism?  And can anyone tell me what "EINSTEIN" stands for?

*Rivaroxaban is FDA-approved for the prevention of stroke in atrial fibrillation and DVT in post-orthopedic surgery patients.

Download the Venous Thromboembolism: Treatment teaching module at

Thursday, March 8, 2012

Supersize Me (and My Liver)

Fatty liver of a 14-year-old.  Source:  AJR
What's the most common cause of newly diagnosed chronic liver disease in the U.S.?

It's not alcohol.  It's not hepatitis C.  It's nonalcoholic fatty liver disease, or NAFLD, which now accounts for 39% of new liver disease diagnoses.

You would think with the epidemic of NAFLD, along with the rising rates of obesity and metabolic syndrome, that there would be an explosion of drug development in this field.  In fact, there have been but a handful of small, randomized, controlled trials.  The NIH clinical trial registration site lists only 193 ongoing or completed trials in NAFLD or NASH (nonalcoholic steatohepatitis, the more inflammatory form of NAFLD), compared to 1370 trials in hepatitis C.

Perhaps part of the reluctance in conducting trials has to do with the fact that simple NAFLD, without NASH, has not been definitively linked to increased liver-related complications or mortality.  NASH can progress to cirrhosis, end-stage liver disease, hepatocellular carcinoma and death, but the only way to distinguish NAFLD from early NASH is by liver biopsy.  At present, most patients who otherwise fit the clinical characteristics of fatty liver disease (elevated transaminases; risk factors such as hyperlipidemia, obesity or type diabetes; and a fatty liver on ultrasound or CT) do not undergo biopsy.  Will that change with the publication of the 2010 PIVENS trial?

PIVENS is the largest well-designed treatment trial of NASH, comparing pioglitazone, vitamin E, or placebo in 247 patients.  Notably, this trial excluded diabetics.  Since insulin resistance and oxidant damage are thought to be involved in the pathophysiology of NASH, it makes sense to study an insulin sensitizer and an antioxidant.  Patients were randomized to one of the three treatment arms for 96 weeks, then re-biopsied.  The primary outcome was improvement in histologic features.

Both the pioglitazone and vitamin E groups had significant drops in their transaminase levels.  Only the vitamin E group, though, experienced a highly statistically significant histologic improvement -- 43% for vitamin E, compared to 19% for placebo, with a number needed to treat of about 4.  Pioglitazone showed a trend towards improved histologic changes, but it didn't make the cut-off for statistical significance.  The investigators were no doubt disappointed in this finding, because they then took pains to explain that the pioglitazone group had less hepatocellular ballooning (a marker of more active disease) at baseline.  An imbalance in baseline characteristics is a known hazard of small clinical trials, and many researchers would have simply smacked their heads with a "D'oh!" and called it a day.  Instead, the PIVENS group went back and performed some post hoc data massaging, and found that if they had used slightly different histological criteria, pioglitazone would have been superior to placebo.

Basing clinical decisions upon subgroup analysis, especially post hoc (i.e., after the fact) analysis, is fraught with danger.  But even if we took these results at face value, should we be prescribing pioglitazone or vitamin E to our NASH patients?

First, note that these were biopsy-proven NASH cases.  Extrapolating the results to patients with NAFLD would not be appropriate.  Even if these drugs were effective in this population, the number needed to treat would be much higher, since many patients with NAFLD never progress to a more inflammatory stage.  Also, note that PIVENS excluded those with diabetes -- patients to whom you might prescribe pioglitazone anyway.  Would you be willing to give pioglitazone indefinitely to a non-diabetic with NASH, especially without a proven histological benefit?  As for vitamin E, it may be tempting to throw this pill at all your fatty liver patients, since it's cheap and "only a vitamin."  Just note that anti-oxidants, at least in pill form, have failed to live up to their promise in most clinical trials, and one meta-analysis even found that vitamin E increases overall mortality.

Finally, there haven't been any studies large enough to detect a difference in clinically significant outcomes, such as complications of cirrhosis.  One trial in progress is randomizing 1500 patients with NASH to either a polypill containing aspirin, a statin, hydrochlorothiazide and enalaparil or no intervention, with a 5-year follow-up.  Even with that many subjects, it still doesn't have the power or the duration to detect a difference in liver-related morbidity. 

Oh, and one more catch:  If you want to sign your patient up for that study, he'll have to fly to Tehran.  What an irony that the Axis of Evil is working on something that may end up benefiting our citizens even more than their own.

Friday, March 2, 2012

Experimenting on the Unconscious

Those of you who read the last post about the RAMPART study, comparing IM midazolam to IV lorazepam in status epilepticus, may have wondered: How did the investigators obtain informed consent from the patients?  The answer is, they didn't.  At least, not until after the patient had already received the study medication.

There is no way to get prospective consent from an unconscious patient.  Even if a family member is present, time is of the essence when treating status.  The paramedics don't have the time to discuss the risks and benefits of the study, and families can't be expected to make a snap decision under duress.

This doesn't mean, though, that researchers are allowed to experiment willy-nilly on unconscious or confused patients.  The U.S. Food and Drug Administration allows exceptions to informed consent requirements under the following conditions (among others):
  • Life-threatening condition requiring urgent intervention
  • Unproven or unsatisfactory treatments
  • Obtained informed consent is not feasible
  • No reasonable way to prospectively identify individuals who may become eligible
  • Possible direct benefit to study participants
Examples of patients who might be eligible for such a study include trauma or cardiac arrest victims.

Although the FDA tries to set clear criteria for what kinds of studies qualify for this exception, they leave room for interpretation.  For example, you can't really say that the standard treatment of status epilepticus is "unproven" or "unsatisfactory."  The majority of patients will respond to anticonvulsants, though some may require aggressive ICU support.  In addition, while a study may hold the promise of benefit to study participants, there is no guarantee against possible harm.

Perhaps the most notorious trial to meet the exemption from informed consent was the PolyHeme study, which randomized trauma patients to synthetic hemoglobin or crystalloid solution in the field.  While there was no overall difference in mortality, those in the PolyHeme group had significantly more adverse events, including myocardial infarction.

Even touchier, studies that qualify for the exemption from informed consent are bound to enroll more racial minorities, because of their higher rates of trauma or use of emergency medical services.  For example, one study found that blacks represented 41% of patients on a U.S. trauma registry, even though they comprised only 20% of the population in those areas. Moreover, the majority of trauma patients enrolled in a hypertonic saline trial, which waived prior informed consent, were black.  On a reassuring note, blacks who met the study eligibility criteria were not enrolled disproportionately more than other racial groups.  As for RAMPART, the authors noted in their results that "the overall number of subjects who were black reflected the proportion of blacks in the subject population from which the sample was drawn."  Whether this is enough to allay the suspicions of a community that has suffered from unethical experimentation in the past remains to be seen.

There is one way that patients can refuse to participate in such a trial.  In RAMPART, paramedics did not enroll patients who wore a medical-alert tag that read "RAMPART declined" -- but how would a patient know to wear such a tag in the first place?  The investigators are responsible for publicizing the trial, for example, in YouTube videos or newspaper ads.  Given that only about 400 patients have viewed RAMPART's video at the time of this posting, my guess is that most people, even those with seizure disorders, were unaware of the study or the fact that they could opt out.  It also seems unfair to place the burden on patients to wear these bracelets 24-7 just to avoid possible, unwanted experimentation.  And what if you want to avoid participating in all the emergency trials taking place in your area - must you don multiple tags or bracelets?

Despite these concerns, I still think RAMPART was an ethical study, and here's the reason why:  Even before the commencement of this trial, IM midazolam, the experimental treatment, was fast becoming the drug of choice among many EMS providers, due to its ease of storage and administration.  Surely performing a randomized trial of this drug in unconscious patients is more ethical than continuing to give those same patients a medication of uncertain efficacy.

Thursday, February 16, 2012

STAT Treatment of Status

The lead article in today's New England Journal of Medicine is a randomized, double-blinded trial of IM midazolam (Versed) vs. IV lorazepam for the prehospital treatment of status epilepticus.  I don't know how many paramedics read this blog, but ER docs, hospitalists and neurologists will need to be familiar with RAMPART (Rapid Anticonvulsant Medication Prior to Arrival Trial), so let's review it.

The study was designed to see if IM midazolam is noninferior to the current standard of care, IV lorazepam.  While lorazepam has been shown to be an effective initial treatment for status, it's not always easy to get IV access, especially while a patient is actively seizing.  Lorazepam also has to be refrigerated, unlike midazolam.

Adult or pediatric patients who had been continuously convulsing for 5 minutes, or those with intermittent seizures without regaining consciousness for 5 minutes, were eligible for the study.  Patients with certain acute precipitants of seizures, such as cardiac arrest, were excluded.  The paramedics tore open a study box, which activated a voice recorder.  The box contained an intramuscular auto-injector and a preloaded intravenous syringe.  One contained the active drug, while the other contained placebo.  The paramedics were instructed to give the IM medication first, then establish venous access and give the IV medication next.  During this time, EMS personnel verbally recorded when they gave the medication, and when the seizure stopped.

The primary outcome was the cessation of seizures before arrival to the ED, without the need for rescue medication.  Even though the study was powered for non-inferiority, IM midazolam ended up being superior to IV lorazepam: 73% in the midazolam group stopped seizing, compared to 63% of the lorazepam group.  The absolute risk reduction was 10%, with a respectable number needed to treat of 10.  There were no differences in the secondary outcomes of intubation or seizure recurrence.

RAMPART was randomized, double-blinded, intention-to-treat, with excellent follow-up (it's hard to lose an unconscious patient in an ambulance).  My only beef is that it seemed designed to give IM midazolam an edge:  the paramedics were instructed to give the IM drug before the IV one.  The trial could have randomized the order of medication given.  And indeed, the mean time to delivery of active treatment was significantly shorter in the IM group (1.2 minutes) than the IV group (4.8 minutes).  On the flip side, once the medication actually got into the patient, lorazepam was faster in terminating the seizure (1.6 minutes) than midazolam (3.3 minutes).

So is midazolam superior to lorazepam in the treatment of status, or is giving an IM medication just faster and thus superior to starting an IV?  We may never know.  The authors of this paper appropriately titled it "Intramuscular versus Intravenous Therapy..." instead of "Midazolam Is Superior to Lorazepam...."   Note that this study isn't applicable to patients who already have an IV.  If a hospital inpatient goes into status, it makes sense to give IV instead of IM treatment because of its faster onset.

Next post:  RAMPART was extremely well-designed.  But was it ethical?

Wednesday, February 15, 2012

Amoxicillin Effective for Sinusitis? It 'Snot.

How many studies does it take to prove a lack of clinical benefit?  There have been multiple randomized, controlled trials of antibiotics for acute, uncomplicated sinusitis, and meta-analyses have generally shown only a marginal benefit.  That's because the majority of sinusitis cases resolve spontaneously, given enough time.

So what's different about the latest RCT published in this issue of JAMA?  The authors of this study decided to test the CDC's conservative recommendations for antibiotic treatment, enrolling only patients who had moderate to severe symptoms present for at least a week.  (You could also be enrolled if your symptoms worsened within less than a week.)  Physicians diagnosed patients using clinical, rather than imaging, criteria, which closely mirrors what is done in the community.  Participants were randomized to either 10 days of amoxicillin or placebo.  A narrow-spectrum antibiotic was chosen, again in line with the CDC's recs, and at the time the study was performed, S. pneumoniae antibiotic resistance was rare.

The primary outcome was improvement in rhinosinusitis symptoms by day 3, as measured by the SNOT-16 scale.  (Hey, amoxicillin is a generic.  You're not going to find any sexy acronyms in this paper.)  The scale looked at 16 symptoms, rating them from 0 (none) to 3 (severe).  Both groups were allowed symptomatic treatment with pain relievers and decongestants.

One-hundred and sixty-six adults were enrolled.  The mean score on the SNOT-16 scale decreased from 1.7 to 1.1 in both groups by day 3.  There was also no difference in outcomes by day 10, with both groups dropping to 0.5.  There was a statistically significant decrease in symptoms in the amoxicillin group on day 7, but not a clinically significant one: the absolute difference in the symptom score was only 0.2.

Sinusitis accounts for 1 out of 5 antibiotic prescriptions in the U.S.  This well-designed trial confirms that they have but a borderline benefit, even in those with prolonged, moderate to severe symptoms.  Will primary care doctors finally listen to the results of this latest study?  

More importantly, will our patients?

Tuesday, February 14, 2012

Chocolate and the Happy Heart

In honor of Valentine's Day, one of our hospitalists, Tom Ormiston, gave a special "Bite-Sized EBM" presentation to the residents, covering a number of recent, thought-provoking papers.  One of them was a meta-analysis of chocolate consumption and the risk of cardiovascular and metabolic outcomes.  Before proceeding further, we must disclose a potential conflict of interest:  Dr. Ormiston passed out chocolate hearts to the audience (although no chocolate-sponsored ballpoint pens), and they were divine.

The meta-analysis included seven large observational studies of chocolate consumption.  The investigators chose to examine cardiometabolic outcomes, because of cocoa's purported "antioxidant, anti-hypertensive, anti-inflammatory, anti-atherogenic, and anti-thrombotic effects."  The analysis compared the participants who reported the highest level of chocolate consumption to those with the lowest. Here were there results:
Should this be called a Black Forest plot?

Remember, a relative risk (RR) less than one means that one group (in this case, the high chocolate group) has a lower rate of events than a comparison group (the low chocolate group).   The high chocolate group had an overall 37% decrease in the risk of cardiovascular disease, a 29% decrease in stroke, and no difference in heart failure.  Only one study, performed in Japan, looked at the risk of diabetes and found that there was a trend towards decreased diabetes in the high chocolate group.  The investigators had planned a subgroup analysis of dark vs. milk vs. white chocolate, but unfortunately, none of the studies collected this data.

All of you EBM acolytes can go ahead and start poking holes in this review:
  • All the studies were observational, and we know that when it comes to meta-analysis, garbage in,  garbage out.  In their defense, most of the studies were higher quality and controlled for confounders, such as body mass index, other dietary factors and education level.  Still, what about the intangibles -- are the lovelorn less likely to receive chocolate, and thus more likely to have a bad cardiovascular outcome?
  • The studies reported chocolate consumption using different scales.  "High" chocolate consumption could mean anything from at least once a week to more than once a day.
  • Some of the studies included only healthy participants, while others included those who already had a cardiovascular diagnosis.  Is chocolate useful only for primary prevention, or can it be used for secondary?
Before we become shills for Willy Wonka, let's recall the disappointing results of anti-oxidant vitamin trials, which held initial promise but may actually increase overall mortality.  And unlike typical dietary supplements, cocoa is usually consumed in high-fat, high-sugar form, so any recommendations for the consumption of chocolate would be premature.  The authors of this meta-analysis concluded appropriately, "Considering the limited data available, any conclusions should be cautious....Corroboration is now required from further test causation rather than just association."

But since it's Valentine's day, we don't wish to rain on anyone's parade.  Feel free to enjoy your chocolate, as long as you take it with a grain of sugar.