Monday, December 22, 2014

Antibiotics, Therapeutic Hypothermia, and Damned Lies (Statistics)

This brief article bubbled back up to the surface the other day, with a colleague claiming an “NNT of 5” for mortality for antibiotic prophylaxis following cardiac arrest.  An NNT of 5 is excellent, mortality is a bad thing – so why aren’t we doing this?

This is a retrospective review of 138 patients resuscitated from OHCA and admitted to a British ICU, all receiving therapeutic hypothermia, stratified by receipt of antibiotics within 7 days of the ICU stay.  Mortality was 56.6% in those not receiving antibiotics and 75.3% for those receiving antibiotics, thus the independent association with antibiotics use and survival with an NNT of 5.

This is, unfortunately, an egregious example of an abstract exaggerating and misleading from the correct conclusion, including the authors own results and discussion.  Any mortality difference between groups occurred in the first 3 days following admission – thus, immortal time bias, as those patients who could receive antibiotics were, by definition, those who lived long enough to receive them.  This fact is ultimately reflected in the presented results, in which the Cox proportional hazards model identified no significant predictive effect of antibiotics on survival.  However, the authors – particularly in the abstract – focus on the superficial mortality observation, and inappropriately promote the potential value of prophylactic antibiotic use.  This supposed NNT of 5 is nonsense, and ought never have been published.

Don’t just read the abstract!

Also, editors of Resuscitation, don’t let authors write such abstracts!

“Early antibiotics improve survival following out-of hospital cardiac arrest”

emlitofnote will be on holiday hiatus until December 29th.  Cheers!

Friday, December 19, 2014

Endovascular Therapy, Unproven Efficacy, Unproven Effectiveness

With the publication of MR-CLEAN two days ago, the medical world (especially Covidien, Stryker, and Penumbra) is ready to throw out all previous neutral trials – MR-RESCUE, SYNTHESIS, IMS-3 – and rush headlong into endovascular therapy for acute ischemic stroke as a new standard of care.

Nonsense, you say?  Not when financial and professional conflicts-of-interest coordinate to drown out the skeptics.

And, frankly, the reality is – despite trials failing to demonstrate benefit, endovascular therapy is already in widespread use.  The underlying tissue-reperfusion hypothesis upon which thrombolytic therapy is predicated is too compelling to wait for proof of efficacy or effectiveness.  Alas.

This is ICARO-3, essentially an ongoing prospective registry of patients considered by experts as optimal candidates for endovascular interventions.  These are all proximal internal carotid artery occlusions, having the among the least favorable rates of recanalization with IV tPA alone.  These authors obtained data on 324 cases between 2010 and 2013 receiving endovascular therapy and 324 matched controls (including 253 from the original ICARO study).  The endovascular cohort included a distribution of patients who received intra-arterial lysis, mechanical retrieval, systemic thrombolysis, or a combination of those treatments, while the tPA cohort received systemic thrombolysis alone.

Patients were generally well-matched on baseline characteristics, with an overall median NIHSS of the entire cohort of 16.  Ultimately, 18.2% of the IV tPA-only cohort had an mRS 0-1 at three months, compared with 20.7% of the endovascular cohort, and OR of 1.17 (95% CI 0.79-1.73).  However, 37% of the endovascular cohort experienced intracranial bleeding of which 6% was fatal, compared with 17.3% and 2.2% in the tPA-only cohort.  In summary, the outcomes – both positive and negative – were a wash.  The authors try to splice out an adjusted subgroup of specific types of endovascular interventions with improved outcomes compared with controls, but these statistical calisthenics are best left unmentioned given their limited validity.

So, until MR-CLEAN, all the randomized trials for efficacy have been neutral.  ICARO-3, a real-world effectiveness observation – also neutral.

Is it too late to derail the bandwagon?

“Intravenous thrombolysis or endovascular therapy for acute ischemic stroke associated with cervical internal carotid artery occlusion: the ICARO-3 study”

Wednesday, December 17, 2014

MR-CLEAN & the New Golden Age

I, among many others, have been highly skeptical of thrombolytic therapy and its role in the treatment of acute ischemic stroke.  As has been well-documented, a few trials were positive, many were neutral, and a few were stopped early for harm or futility.  To most of us, this indicates a therapy for whom only a small subset of those treated are ideal candidates for benefit, and the margin between benefit and harm is razor thin.

In my previous posts, I’ve sighed wistfully at the hope of The Next Big Thing in stroke treatment – local endovascular therapy, akin to percutaneous coronary intervention.  However, each major endovascular trial published in the New England Journal last year failed to demonstrate benefit.

MR-CLEAN is different.  MR-CLEAN is rather unambiguously positive.  To be zero or minimally disabled?  The endovascular intervention is favored 12% to 6%.  “Functionally independent”, a modified Rankin Scale of 0-2, favors endovascular intervention 33% to 19%.  A number needed to treat of, apparently, ~8 for independence is nothing to scoff at.

But why?  It’s very similar to IMS-3, which was stopped early due to futility.  Patients are about the same age.  The comparator – usual care, typically tPA – is the same.  Median NIHSS is about the same.  The differences are quite subtle.  Patients were randomized earlier in IMS-3 compared with MR-CLEAN, with the implication IMS-3 includes patients whose natural course was superior, whereas MR-CLEAN enrolled "non-responders".  The other difference, and the one you’ll hear by far the most frequently, is that MR-CLEAN utilized modern stent retrievers, rather than such killing machines as the MERCI device.  Newer, as you've always been taught, is better.

But, clearly, there’s something else we simply cannot splice out of these data.  Patients in MR-CLEAN did awful.  Recall NINDS, where a tPA cohort with a median NIHSS of 14 resulted in 39% attaining mRS 0-1.  In IMS-3, intravenous tPA with a median NIHSS 16 resulted in 26% mRS 0-1.  In MR-CLEAN, intravenous tPA with a median NIHSS of 18 resulted in 6% mRS 0-1.  Patients in MR-CLEAN did recanalize at a greater rate than those in IMS-3, 58% vs. 23-44%, owing to the improved performance of modern retrievers.  In a world where definitively opening the vessel, where reperfusion means time=brain, this makes sense.  But, like NINDS, the positive results do not seem so much to result from the intervention, but rather from the control group simply doing unwell.

As the embargo lifts, I’m sure this post is one of a tiny minority wondering if this is fool’s gold.  If you think of p-values like likelihood ratios, as initially intended, the presence of multiple prior neutral evaluations makes the bar for success that much higher in follow-up trials.  These are excellent results, results I’d like to believe in, but the totality of evidence to date requires they be validated.

I wholeheartedly expect they will not.  Prepare for the full onslaught of hype regarding endovascular therapy for stroke.

“A Randomized Trial of Intra-arterial Treatment for Acute Ischemic Stroke”

Monday, December 15, 2014

Not Osletamivir, But Peramivir

Because the CDC is still pounding the pavement hawking neuraminidase inhibitors, the market is ripe for profit.  Not content to allow Roche all the fun, BioCryst Pharmaceuticals is trying to edge its way into the game.

BioCryst, as you now know, is the manufacturer of peramivir, a neuraminidase inhibotir whose initial claim to fame was emergency approval for intravenous use during the initial H1N1 pandemic.  They were in the news a few months back when they presented results for combined trials highlighting its use as a single intramuscular dose in the outpatient setting.  Now, the results of said trial are fully published, comprising combined populations of a phase 2 and under-enrolled phase 3 trial.

What net benefit, in these sponsored trials, in the lab-confirmed influenza population?

About an 18-hour reduction in symptoms, from a mean of 148 hours down to a mean of 130 hours with the 300mg IM injection.  The number of days until resumption of usual activities was nearly identical, from a mean of 9.4 in the 300mg cohort to 10.2 days in the placebo cohort.

And, what’s interesting, if you go back to some of the earlier studies, you might find nearly all the observed benefit associated with antiviral treatment was observed in the cohort treated within 24-hours, with no clinically relevant benefit between 24-48 hours.  However, you’ll only rarely see such detail presented, as it diminishes the viability of the product.

And, finally, just to add to your peramivir update – a clinical trial regarding the value of intravenous peramivir in hospitalized patients with influenza.  In a rather complicated and confounded design with only a few hundred patients, no benefit was demonstrated by use of peramivir whether as NAI monotherapy or adjunctive therapy to other neuraminidase inhibitors.  In the 121 patients who received peramivir monotherapy as part of the standard of care, there was no difference in any secondary infection or mortality.

But, I’m sure peramivir will rapidly be added to our antiviral armamentarium as a lovely money-wasting enterprise, regardless.

“Single dose peramivir for the treatment of acute seasonal influenza: integrated analysis of efficacy and safety from two placebo-controlled trials.”

“Efficacy and Safety of Intravenous Peramivir for Treatment of Seasonal Influenza Virus Infection”

“Evaluation of Intravenous Peramivir for Treatment of Influenza in Hospitalized Patients”

Friday, December 12, 2014

All Falling Apart for Epinephrine

A staple of resuscitative care, epinephrine is enshrined in the ACLS algorithms for the pulseless patient.  And, what seems to be clear – flogging the heart with vasoactive agents produces, at least, a temporary physiologic response.  Unfortunately, such sympathomimetic abuse results in multiple adverse effects, included coronary and cerebral vasoconstriction – and the net effect, perhaps, is negative.

This is a retrospective review of 1,556 patients admitted to a “cardiac arrest center” in Paris, France, evaluating cerebral performance outcomes depending upon pre-hospital administration of epinephrine.  Of the 1,134 patients who received epinephrine, 17% ultimately had good outcome.  Of the remaining 422 patients who did not receive epinephrine, 63% ultimately had good outcome.  The authors also demonstrate worsening outcomes for epinephrine administration in a dose-dependent fashion.  Cheers.

Of course, it should follow naturally the patients receiving epinephrine probably did so because they were judged to need it – whereas, contrariwise, those not receiving epinephrine probably did not receive it because they were sufficiently stable.  And, then, failure to achieve initial return of spontaneous circulation begets additional doses of epinephrine.  Thus, you have a selection bias in which sicker patients were naturally allocated to epinephrine and less-sick patients to non-treatment – and an accounting for the dose-response relationship.  The authors perform multiple adjustments and propensity matches in an attempt to prove the outcome disparity durable to these validity challenges, but the astute reader may find them insufficient.

This study, nor any of the observational studies preceding it, definitively prove the harms of use of epinephrine pre-hospital outweighs the benefits.  I have, at least, stopped routinely using epinephrine in undifferentiated cardiac arrest, and rather try to select specific patients for whom the underlying etiology seems appropriate.

“Is Epinephrine During Cardiac Arrest Associated With Worse Outcomes in Resuscitated Patients?”

Wednesday, December 10, 2014

BLS is More

Coming from JAMA Internal Medicine’s “Less is More” series is this latest evaluation prehospital care for out-of-hospital cardiac arrest.

These authors compared basic life support – CPR and defibrillation – with advanced life support – as above, but with feeling (and intubation and pharmacologic flogging) – to determine, once and for all, the winner.  And the winner, by a wide margin, is BLS – 13.1% vs. 9.2% survival, and 21.8% vs. 44.8% poor neurologic functioning.  Ah, so.

Naturally, as with everything in life, the details are far muddier than this simple conclusion.  The study population established BLS or ALS level of care by utilizing a Medicare procedural coding database for ambulance services, not specific patient-level interventions.  Patients were then linked to a Medicare billing database using ICD-9 codes for cardiac arrest “present on admission”, with multiple exclusions owing to poor underlying data quality.  And, finally, their primary outcome of neurologic performance was inferred from ICD-9 codes for brain injury and vegetative state, not direct observations.  To say these authors have performed potentially specious and indirect measurements is quite the understatement.

All that considered, however, these observations very likely reflect underlying truths.  Any advantage of pre-hospital ALS over BLS for OHCA remains steadfastly unproven.  Such interventions as endotracheal intubation and sympathomimetic therapy are likely harmful, despite having been persistently assumed to be beneficial.  And, as this article points out, ALS-trained professionals are far more expensive than BLS – further salting the wound.

More, almost certainly not better.

“Outcomes After Out-of-Hospital Cardiac Arrest Treated by Basic vs Advanced Life Support”

Monday, December 8, 2014

The "Golden Hour" of tPA

As part of every Genentech-sponsored CME or medical school curriculum presentation on thrombolysis in acute ischemic stroke, you see a graph like this:

This is the “time is brain” mantra, where every supposed passing second without flow destroys another mass quantity of brain cells.  The theory: on the edge of the infarcted brain tissue, there remains yet a thin rim of cells between vascular territories, just barely hanging on.  Timely reperfusion saves these cells – and tPA increases recanalization rates incrementally beyond control, in certain patients, and in certain vessels.  However, again, in certain patients, and in certain vessels, the tPA results in serious intracranial bleeding.  This risk/benefit trade-off remains a cornerstone controversy in Emergency Medicine and Neurology, owing to the paucity of unbiased clinical trial data.

But, most patients eligible for tPA arrive at the hospital far at the low end of the curve – where the time-dependent effects are weakening.  As such, the new magic is to take the tPA to the patient – delivering thrombolysis in a van down by the river.  The STEMO from Germany, and its PHANTOM-S project, paved the way for other “mobile stroke units”.  This report aims to evaluate the “Golden Hour” of stroke thrombolysis – patients receiving tPA within 1-hour of symptom onset – where the time-is-brain ought wholeheartedly manifest.

Ah, but, so – such an effect, unfortunately, was not conclusively observed.  Comparing 78 patients who received tPA in fewer than 60 minutes (median 50 minutes) with 451 patients receiving tPA in greater than 60 minutes (median 105 minutes), these data support no useful conclusions regarding the effectiveness of such timely delivery.  The unadjusted analysis shows no difference in outcomes between groups, but, alas, the groups are grossly imbalanced.  Adjusted analyses, in their crude adjustments, tend towards benefit in the <60 minute cohort, but such post-hoc comparisons can only be considered exploratory.  Of course, the accompanying editorial “Prehospital Thrombolysis for Stroke: An Idea Whose Golden Hour Has Arrived” cleanly ignores the adjusted nature of these data, and definitively endorses the observed benefit.

Considering pre-hospital thrombolysis provides such excellent opportunity to truly test the time-is-brain hypothesis, it’s a shame its proponents are not taking such an uncritical view of the opportunity for study.

“Effects of Golden Hour Thrombolysis: A Prehospital Acute Neurological Treatment and Optimization of Medical Care in Stroke (PHANTOM-S) Substudy”

Friday, December 5, 2014

FFR(CT) Is Back! And Better Than Ever!

Invasive coronary angiography is problematic – specifically, it’s invasive.  Radial artery approaches have reduced the incidence of bleeding complications, but it remains a costly and non-risk-free procedure.  In lieu of ICA, CT coronary angiography has become increasingly popular.  However, CCTA is problematic – specifically, it’s inaccurate.

A few years ago, DeFACTO was published in JAMA and covered on this blog, a study evaluating a non-invasive model of fractional flow reserve added on to CCTA in an attempt to improve accuracy at identifying true culprit lesions.  DeFACTO was negative – specifically, the per-vessel performance at predicting flow-limiting lesions compared to the traditional 50% stenosis cut-off of CCTA was nearly identical.

Two years have passed, however, and we have a new study – NXT – using the next iteration of the HeartFlow software, and, of course, performed by authors with robust conflicts-of-interest.  Now, improvements in image quality and luminal modeling – as well as refined exclusion criteria to prevent troublesome images confounding their software – have improved performance to the point where, yes, it now seems to out-perform baseline CCTA.

The catch, of course, is the CCTA criterion standard is abysmal.  Compared with the ACRIN-PA or ROMICAT studies with their pro-CCTA COI, in which CCTA is the best thing since sliced bread, these folks are unconcerned with the collateral damage of degrading CCTA.  In this study, as performed on patients with suspected CAD, of 237 vessels read as “positive” by CCTA (>50% stenosis), only 83 (35.0%) were actually judged to be flow-limiting lesions on ICA – which is to say, false positives doubled the true positives.  Likewise – in contrast to the ROMICAT and ACRIN studies purveying CCTA as a bulletproof mechanism for discharge – 17 of 247 (6.8%) patients read as negative by CCTA (<50% stenosis) actually had flow-limiting disease.

False positives more two-thirds of the time?  And then a 7% miss rate of clinically important stenosis?  Basic, anatomic CCTA as previously described – not as fantastic as you've been led to believe.

The HeartFlow software?  Perhaps.  Effectiveness evaluations absent pervasive COI will be necessary to truly describe its value.

“Diagnostic Performance of Noninvasive Fractional Flow Reserve Derived From Coronary Computed Tomography Angiography in Suspected Coronary Artery Disease”

Wednesday, December 3, 2014

4-Factor Works for Factor Xa Inhibitors

The newest study regarding the reversal of the novel oral anticoagulants also concerns the newest of their family – edoxaban, joining a market already occupied by rivaroxaban and apixiban.  Yes, it’s just another “me too” drug trying to shoehorn its way into the massive warfarin-replacement market – but, at least, this is useful evidence.

4-Factor prothrombin concentrate complexes have been established as a treatment option in the setting of hemorrhagic complications during anticoagulation with the Factor Xa inhibitors.  However, the initial studies of PCCs primarily concerned themselves with laboratory markers of reversal, and any live bleeding studies were in animals (rabbits!).  For this study, however, the authors & Daiichi Sankyo paid 110 healthy volunteers to take edoxaban, then undergo punch biopsy, and then receive either placebo or 4-factor PCC in either 10 IU/kg, 25 IU/kg, or 50 IU/kg doses.

And, pleasantly, it works – in a dose dependent fashion, even, with 50 IU/kg providing the most effective reversal.  So, this essentially confirms what we thought we knew about PCCs based on surrogate markers and animal studies.  These volunteers only suffered a minor dermatologic procedure – and are not critically ill patients with other associated coagulopathies – but it still provides reasonable insight.

How this reversal strategy will fit versus andexanet alfa, a NOAC-specific reversal agent, remains to be seen – particularly with regards to cost and pro-thrombotic adverse effects.

“Edoxaban Effects on Bleeding Following Punch Biopsy and Reversal by a 4-Factor Prothrombin Complex Concentrate”

Monday, December 1, 2014

More Futility for Mechanical CPR

A few weeks ago we reported on the use of the LUCAS-2 for crushing the thorax, resulting in significant internal injuries.  None of these injuries were judged to have contributed to any patient’s demise – but, still, concerning.  This could be more forgivable if there were other advantages, say, survival.

Unfortunately, yet again, mechanical CPR fails to demonstrate superiority.

This is the PARAMEDIC trial, from the Lancet, a cluster-randomized trial aiming to demonstrate the superiority of the LUCAS-2 device as compared to manual CPR.  Based on a deployment of 143 available LUCAS devices, these authors compared 1,652 patients randomized to attempt device deployment compared with 2,819 in a control group.  There were no important differences in the baseline characteristics of the two groups, and other measures of initial treatment were similar.  And, in the end, there was no difference between initial return of circulation, event survival, 30-day survival, 3-month survival, or 12-month survival (all dismal).

The mildly interesting aspect to this trial is that it was not directly a test of the LUCAS device – but, rather, a pragmatic trial comparing deployment with LUCAS as an option with deployment without.  Thus, 638 patients in the clusters randomized to include the LUCAS device received manual compressions.  However, on multiple causal effect analyses, no advantage was detected from cases in which the LUCAS device was used, nor was a selection bias apparent.

In the absence of efficacy data, use of mechanical devices is defended primarily on the basis of convenience – freeing up hands for other tasks.  However, this argument fails in the face of these data – as it does not appear the handsfree advantage, when available, results in improved patient outcomes.

“Mechanical versus manual chest compression for out-of-hospital cardiac arrest (PARAMEDIC): a pragmatic, cluster randomised controlled trial”