Friday, February 27, 2015

The Case of the Bloody Lumbar Punctures

Modern evaluation for aneurysmal subarachnoid hermorrhage, with some debate, may include definitive non-contrast CT performed within six hours of symptom onset.  The traditional evaluation, and still recommended beyond six hours, involves a lumbar puncture, looking for red blood cells or xanthrochromia.

This latest tale of woe from Jeff Perry’s SAH data details the pragmatic effectiveness of the traditional pathway, focusing on the primary confounder: traumatic taps.  They report on 1,739 patients undergoing lumbar puncture as part of this evaluation, and, unfortunately, the numbers are grim:  641 (36.8%) samples were abnormal in the final tube of CSF collected.  However, it isn’t so bad – 476 of those had fewer than 100 RBCs x 10^6/L, with many having only a handful of cells.  But, still, that leaves 165 patients with fairly substantial numbers of RBCs in their CSF.

Because, all told, only 15 received a final diagnosis of aneurysmal SAH.

Why is this so grim?  Because 419 of these 626 patients with RBCs on their LP subsequently were subjected to angiography – with 404 of them negative.

And xanthrochromia?  Some predictive value – 7 of 15 patients diagnosed with SAH displayed xanthrochromia, but, obviously there were 8 patients with SAH who did not, along with 16 instances of xanthrochromia in patients without SAH.

The final gist of the paper is to generate a 100% sensitive cut-off to exclude SAH – for which the authors choose 2000 x 10^6 and absent xanthrochromia.  This results in a specificity of 91.2% and a positive LR or 11.4.  This is a pretty good positive LR, but, unfortunately, given such a vanishingly rare disease, the PPV was only 21.4% in their cohort.

However, one major flaw in this study is it doesn’t usefully describe the population of true interest to Emergency Physicians – the test characteristics of those with a negative CT and a positive LP.  There were 77 patients who did not undergo CT prior to LP, but, more importantly, 10 of the patients included in this cohort had visible SAH on CT recognized by the staff radiologist, but not the Emergency Physician.  Therefore, if you practice in a setting without neuroradiology coverage, this is generalizable.  Otherwise, we can exclude those 10 cases and boggle at the massive resource utilization in terms of LPs and angiography in order to pick up just 5 cases of occult aneurysmal SAH.

In patient-oriented terms – based on these data – the risk of SAH after a negative CT performed greater than 6 hours after onset is about 1 in 330.  Using their cut-off of 2000 x10^6, the chance of a true positive LP is about 1 in 12.  A vast improvement, to be sure, but probably still not a pathway very many patients are going to choose when presented with these odds.

“Differentiation between traumatic tap and aneurysmal subarachnoid hemorrhage: prospective cohort study” (free fulltext)

Wednesday, February 25, 2015

Rampant Underreported Research Misconduct

It is not surprising to hear clinical trials sometimes struggle with data integrity and quality issues.  Such undertakings can be logistically challenging, and certainly any substantial scope of effort leads to the occasional cutting of corners.

However, there are also millions (or billions) of dollars in revenue, along with multiple professional reputations, at stake.  This creates fertile territory for the more nefarious sort of data corruption.  In some instances, the Food and Drug Administration performs site monitoring as evaluation for misconduct.  And, as this study indicates, the FDA sometimes discovers serious issues – issues almost always swept under the rug.

Using a variety of methods, including Freedom of Information Act requests, site exploration, and other FDA published warnings, these authors compiled a list of 421 serious irregularities identified by FDA audit.  However, heavily redacted language in many of the documents discovered precluded linkage to clinical trials – resulting in only 57 published trials that could be linked to serious violations.  These 57 trials resulted in 78 identifiable publications – only 3 of which mentioned or addressed the issues raised by the FDA.  Those three specifically noted data excluded due to protocol errors, data falsification, or inappropriate monitoring.  The remaining 75 publications did not.

A couple examples:
  • 8 of 16 FDA inspections of sites for RECORD 4, a rivaroxaban trial for DVT prophylaxis, identified unblinding, falsification of records, and randomization improprieties.  The associated study publications do not mention such issues.
  • A clinical site in China falsified data regarding apixiban in ARISTOTLE.  Excluding such data from the final study report would eliminate any apparent mortality benefit, but publications continue reporting mortality benefit analyses based on the entire data set.
The lack of transparency and apparent action regarding what is certainly just the tip of the iceberg is staggering.  How is it our own drug safety organization fails to protect patients on such a scale?  Is it any wonder so few clinical trial results hold up on re-examination?

“Research Misconduct Identified by the US Food and Drug Administration”

Monday, February 23, 2015

Sore Throats More Dangerous than Terrorists

1 in 5 Sore Throats Tied to Scary Bacteria, Study Finds

Thanks, HealthDay.

Now, clickbait headlines aside, what on earth are they referring to?  And, if ~50% of acute pharyngitis already receives inappropriate antibiotics – how can this sort of news release help us maintain any sort of reasonable antibiotic stewardship while also cultivating some modicum of Press Ganey approval?

This is a cross-sectional survey of throat swabs performed on a convenience sample of students at the University of Alabama.  PCR for an expanded cohort of bacterial pathogens was performed on 312 patients presenting to the student health center with acute pharyngitis, and compared with 180 asymptomatic volunteers.  Among symptomatic patients, 20.5% of PCR detected Fusobacterium necrophorum, compared with 9.4% of the asymptomatic cohort, leading to such conclusions as: “approximately 11% of cases of pharyngitis in patients coming to this university health clinic were caused by F. necrophorum.”  Group A Streptococcus, by comparison, was detected in only 10.3% of symptomatic patients and 1.1% of asymptomatic.

So, an epidemic of F. necrophorum?  Should we, as these authors suggest, be considering penicillin for all sore throats – based on the feared complication of Lemierre syndrome – regardless of rapid streptococcal antigen testing?


While, we shouldn’t forget about F. necrophorum, particularly in the adolescent/young adult population, it is important to remember the vast majority of pharyngitis – even bacterial – is self-limited, with minimal benefit from antibiotics, either for symptom relief or suppurative complications.  For Group A Streptococcus, it is reasonable to suggest well over 200 cases must be treated with antibiotics to prevent progressive disease, and rheumatic fever is virtually extinct.  However, we have no similarly useful statistics regarding F. necrophorum – mostly because serious downstream complications are so rare they are still literally one-in-a-million case reports.  Symptomatic management and judicious treatment for acute pharyngitis remains the most appropriate strategy, despite the prevalence of this “scary bacteria”.

“The Clinical Presentation of Fusobacterium-Positive and Streptococcal-Positive Pharyngitis in a University Health Clinic”

Friday, February 20, 2015

Distorted Treatment Effects for Steroids for Pneumonia

This is the second “steroids for pneumonia” trial published in the last few weeks.  The last trial, enrolling 785 patients with community-acquired pneumonia, showed a small – but potentially relevant – reduction in inpatient length-of-stay.  No differences were noted with respect to mortality or treatment failure.

This trial, however, is a bit different.  In an effort to maximize the theoretical mortality reduction associated with steroid use in pneumonia, these authors targeted therapy specifically at those in the highest pneumonia severity risk categories and required a CRP >150 mg/L.  Patients were then randomized to 0.5 mg/kg twice daily of intravenous methylprednisolone or placebo.  The primary outcome was “treatment failure”, which was composed of two definitions – one specifically for early deterioration and one for late deterioration.

At face value – the results are excellent.  There was 31% failure rate in the 59 patients in the placebo group, compared with 13% of the 61 patients in the methylprednisolone group.  Deaths were 10% in the methylprednisolone group and 15% for placebo, and few adverse events occurred in either group – these differences, however, did not reach statistical significance due to the small sample size.

But this trial is essentially noise, full of baseline confounders and inconsistencies.  To start, simply, note each center enrolled, on average, one patient every-other month for eight years – only managing to screen 519 total patients with pneumonia for eligibility over the course of the trial.  This does not reflect a well-executed trial infrastructure.  An excess of 11% of placebo patients were admitted to the ICU, reflecting in part a 20% excess of placebo patients with shock as part of their initial presentation.  Shock and multiple organ failure was the major cause of death in the placebo group, compared with disease progression in the steroid cohort.

Furthermore, 40% of the placebo patients presenting with shock did not receive antibiotics within 4 hours of arrival.  Causative organisms were detected for 51% of the steroid cohort, compared with 30% of the placebo group – with 21% of the steroid cohort having a “respiratory virus” compared with 8% in the placebo group.  Antibiotic use was also odd, with the prevailing choice being ceftriaxone plus levofloxacin, rather than the typical ceftriaxone + azithromycin combination used for CAP.

How this managed to get published in one of the supposedly pre-eminent medicine journals is beyond me.  With only 120 patients, all the substantial absolute differences in baseline characteristics and care heavily distort the overall results.  Mostly, unfortunately, it looks like placebo patients were sicker and received less-adequate initial care – and everything measured in this small trial is suspect as a result.

“Effect of Corticosteroids on Treatment Failure Among Hospitalized Patients With Severe Community-Acquired Pneumonia and High Inflammatory Response”

Wednesday, February 18, 2015

Irresponsible Use of NOACs in End-Stage Renal Disease

Frequent readers may have noted this blog is somewhat skeptical regarding the novel oral anticoagulants, with particular criticism reserved for dabigatran.*  The bleeding risks, particularly for dabigatran, are profoundly increased in renal impairment – while the Factor Xa inhibitors simply do not have sufficient safety data to describe their risks in this population.

So, in dialysis patients with zero renal function – wouldn’t it perhaps be safest to continue using our present, time-tested, warfarin anticoagulation strategy?

This review of the Fresenius Medical Care database of end-stage renal patients on dialysis captured 8,064 patients with non-valvular atrial fibrillation who were initiated on anticoagulation between 2010 and 2014.  During this time, 5.9% of patients receiving new anticoagulation were initiated on dabigatran or rivaroxaban, with the remainder started on warfarin or aspirin.  And, dabigatran or rivaroxaban use increased the incidence of minor bleeding, major bleeding, and bleeding-associated mortality – with relative risk increases ranging from ~1.3 for minor bleeding to ~1.7 for hemorrhagic death.  Even rates for ischemic stroke were low in all groups, and no meaningful protective difference for thromboembolic events was observed.  Small baseline differences between the various anticoagulant cohorts are present, but they are probably clinically unimportant.

More bleeding?  More death?  It seems clear it is not responsible medicine to initiate the NOACs in a dialysis population.

“Dabigatran and Rivaroxaban Use in Atrial Fibrillation Patients on Hemodialysis”

*Disclosure: I provided legal consultation pertaining to dabigatran, with funds paid to my institution.

Monday, February 16, 2015

Inappropriately Promoted tPA "Drip and Ship" Safety

“More community hospitals are giving a powerful clot-busting medication to stroke victims, improving their chances of survival and recovery, new research shows.”
This statement comes from the American Heart Association press release regarding this synopsis of the Get-With-the-Guideline Registry.  Part of this statement is true – more community hospitals are using tPA for acute ischemic stroke.  In this review of 44,667 patients treated with tPA over the past decade, 23.5% received tPA outside of a specialized stroke or academic center.

The second half of this press statement is false.

Patients treated by the “drip and ship” method, as community administration of tPA is described, did not have an improved chance of survival.  Patients treated at community hospitals were younger, had less-severe strokes, and had fewer prior strokes – yet their in-hospital mortality was 10.9%, compared with 9.7%.  Additionally, their rate of symptomatic intracranial hemorrhage was 5.7% compared with 5.2%, and they had 1.8% serious tPA-related complications, compared with 1.6%.  These small absolute differences are magnified when adjustments are made for baseline comorbidities, and, in fact the OR for in-hospital mortality increases to 1.23 or 1.33, depending on the precise statistics pursued.  So, of course, the logical leading sentence of the Discussion is:
“In this study of >40000 patients with acute ischemic stroke treated with IV thrombolysis throughout the United States, drip and ship thrombolysis was …. safe.”
A better leading sentence to their Discussion might rather suggest the “drip and ship” model is, in fact, less safe than typical thrombolysis.  Further, they might better suggest the “drip and ship” model should be curtailed while further investigation into additive risks are performed, or to confirm the effects noted from this somewhat dodgy registry data.  But, these authors focus more on explaining away this inconsistency with their narrative than calling for safer, narrower administration of tPA.  After all, these authors are well-funded by industry – including one affiliated with MGH TeleHealth, providing telestroke-enabled thrombolysis:
Dr Sheth is a member of the Get with the Guidelines (GWTG)- Stroke Clinical Workgroup, and he is a Co-Principal Investigator and Executive Committee member for Glyburide Advantage in Malignant Edema and Stroke-Remedy Pharmaceuticals (GAMES-RP), a phase II–trial to prevent swelling in patients with large stroke, funded by Remedy Pharmaceuticals, Inc. Dr Smith is a member of the GWTG- Stroke Workgroup. Dr Kleindorfer discloses speaking engagements. Dr Fonarow is a member of the GWTG Steering Committee; receipt of research support (to the institution) from Patient-Centered Outcomes Research Institute, and he is an employee of the University of California that holds a patent on retriever devices for stroke. Dr Schwamm is the chair of the GWTG-Stroke Clinical Workgroup, a principal investi- gator of the National Institutes of Health–funded MR WITNESS (A Study of Intravenous Thrombolysis With Alteplase in MRI-Selected Patients) trial of extended window thrombolysis for which Genentech provides supplemental site payments and alteplase free of charge, a member of the international steering committee of the Desmoteplase in Acute Ischemic Stroke (DIAS) 3 and 4 trials of extended window thrombolysis, and the director of Massachusetts General Hospital (MGH) TeleHealth. The MGH provides a broad array of telehealth services to hospitals in New England, including telestroke-enabled thrombolysis. Dr Grau-Sepulveda reports no conflicts.
Perhaps the new ACEP Clinical Policy statement can explicitly address such settings in their "systems in place" language.

“Drip and Ship Thrombolytic Therapy for Acute Ischemic Stroke”

Thursday, February 12, 2015

Christmas Comes Early for Endovascular Therapy in Stroke

Quite literally, in fact, considering the timing of the publication of MR-CLEAN – and, now, the triple fall-out from those results.

Due to the positive findings presented by the MR-CLEAN investigators in December 2014, three other ongoing major endovascular trials used this opportunity to check their results early.  ESCAPE, EXTEND-IA, and SWIFT-PRIME – all products of the Covidien Solitaire FR clinical trial machine – ceased recruitment and looked to see if they’d met statistical measures of efficacy.  Obviously, if you’re reading about all three of these trials here today – presented at the International Stroke Conference yesterday – they all found the results they hoped.  ESCAPE and EXTEND-IA were published simultaneously in the NEJM, while SWIFT-PRIME is still in the manuscript drafting stages.

The more robust of the two trials is ESCAPE, whose original target enrollment was 500 patients based on a primary outcome of “ordinal shift analysis” on the modified Rankin scale.  The important feature of all these new endovascular trials is the eligibility population: proximal, large-vessel occlusions with imaging-based evidence of moderate-to-good collateral supply surrounding a small infarct core.  In this particular trial, it was proximal internal carotid or middle cerebral artery trunk, an ASPECTS of 6 to 10, and 50% or more filling of the local pial arterial circulation.  Interestingly, this trial enrolled patients with symptoms out to 12 hours from onset – and did not require pretreatment with intravenous alteplase before intervention.

Ignoring their “ordinal shift analysis” nonsense that only serves to distort the effect size, the key results that matter are these:  in the 316 patients enrolled, mRS 0-2 was 53.0% in the endovascular cohort and 29.3% in the control cohort.  Deaths were improved to 10.4% from 19.0%, despite a small increase in sICH of 3.6% to 2.7%.  3 patients suffered access site hematomas and one retriever perforated the middle cerebral artery.  Pretty good, frankly, for a cohort with a median NIHSS of 16.  Interestingly, 45 patients underwent endovascular intervention without receiving alteplase, and 58% achieved mRS 0-2 – although 20% died.

EXTEND-IA was a much smaller trial, targeting only 100 patients, with coprimary outcomes of reperfusion and NIHSS improvement at 24 hours.  Again, these investigators targeted proximal occlusions with radiographic evidence of salvageable “ischemic penumbra”.  They stopped at 70 patients, again, because they’d met their own complex statistical criteria for efficacy.  All patients in this trial received alteplase within 4.5 hours prior to endovascular intervention.  Finally, yet again, ignoring their specific primary outcomes, the result of interest:  mRS 0-2 was achieved by 71% in the endovascular group compared with 40% in the alteplase-only group.  Deaths were improved to 9% from 20%, although this did not reach statistical significance owing to the small sample size.  Reperfusion and infarct region growth at 24 hours also favored the endovascular cohort, as did the measures of early neurologic improvement.

There are many tiny oddities worth picking over in these trials – and, no doubt, their data will be picked over for further clues and hypotheses.  SWIFT-PRIME was similarly positive, but those data were not yet available for full review.  And, finally, we will all have to come to terms with the early termination of all these trials based on MR-CLEAN.  The sponsor, obviously, is ready to make endovascular treatment the (profitable) standard of care.  However, the full enrollment from these trials would have provided additional information regarding potentially dangerous subgroups.  One hopes there will be ongoing endovascular registries going forward to identify any such patterns.

The key take-home, however, is endovascular therapy for acute stroke has probably finally arrived.  After a decade-and-a-half of generally failed trials, it seems the devices and patient selection have finally improved to the point of clinical utility.  For patients with collateral flow and one of these accessible lesions, it seems clear this therapy should be provided – and neither time of onset or tPA use matter as much as viable brain tissue.  But the obvious key, as shown in MR-CLEAN, is patient selection – ESCAPE only managed to enroll 1.4 patients per month per center, while EXTEND-IA screened 7,798 stroke patients over two years to come up with 70.  This therapy is very much so not for everyone – though, no doubt, Covidien hopes it will become so beyond the eligibility population identified here.

But, for the first time ever, if I were to have one of these specific types of heavily disabling strokes, this is probably the first advanced stroke intervention I'd willingly choose.

“Endovascular Therapy for Ischemic Stroke with Perfusion-Imaging Selection”

“Randomized Assessment of Rapid Endovascular Treatment of Ischemic Stroke"

Wednesday, February 11, 2015

Icatibant for ACE-Inhibitor Angioedema

Oropharyngeal angioedema can be one of the true Emergency Department airway disasters.  Massive and rapidly progressive edema can engulf all usable landmarks and views, necessitating surgical intervention.  No one enjoys these cases – least of all the patient.

This small trial, replete with heavy sponsor involvement, details the utility of icatibant, a selective bradykinin receptor antagonist, for treatment of ACE inhibitor-induced angioedema.  27 patients were randomized either to icatibant or steroids plus an antihistamine.  The mean times to symptom relief were reduced substantially by use of icatibant – with reported total symptom resolution reduced from 27 hours to 8 hours.

Of course, for the three placebo patients meeting the protocol definition of worsening clinical status, the authors arbitrarily set their time to symptom resolution to 61.8 hours – exaggerating (unnecessarily) the difference measured for the primary outcome.  Finally, bizarrely, 4 of these 27 patients were lost to follow-up – all in the placebo cohort.  What sort of effect this would have on the integrity of the results is uncertain.

But, all such misbehaviors aside, icatibant probably works, along with C1-esterase inhibitor and ecallantide.  However, each use of these medications costs between ~$7,000-$10,000 per administration.  Therefore, restraint is necessary to prevent indication creep – and such medications should not be given to all perioral angioedema presentations, and be reserved only as a final option to fend off impending upper airway obstruction.

“A Randomized Trial of Icatibant in ACE-Inhibitor–Induced Angioedema”

Monday, February 9, 2015

The Fixed-Ratio Massive Transfusion Answer

After years of wondering and wandering, we finally have the definitive answer for how best to resuscitate the severely-injured trauma patient – transfusion ratios best mimicking whole blood.  You know, just as we all expected, just as these authors hoped, and just what's been reported from prior observational series and military combat experience.

More or less.

Regardless, this study – the Pragmatic, Randomized, Optimal Platelet and Plasma Ratios (PROPPR) trial – was a remarkable undertaking in logistics.  Each participating Level 1 trauma center incorporated sealed coolers into their ready blood supply, providing a random allocation of product ratio when the massive transfusion protocol was activated.  As transfusion continued, more coolers with the same ratio arrived.  As best as can be implemented, this reduces the immortality bias seen in other observational series – where survivors were survivors in part, basically, because they survived.

This trial randomized patients to 1:1:1 vs. 1:1:2 – that is, equal numbers of FFP, platelets and RBCs, or half as much FFP and platelets as RBCs.  Ultimately, it didn't precisely test those same ratios, except as the initial resuscitation strategy.  Following the intervention period, the 1:1:2 arm caught up a bit with plasma and FFP – but the quantities transfused were not substantial.

Technically, this is a negative trial – the mortality advantage favoring the 1:1:1 cohort did not reach statistical significance at 24 hours or 30 days.  However, the authors powered the study expecting a 10% mortality advantage – and instead it was only 4.2% (95% CI -1.1 to 9.6) and 3.7% (95% CI -2.7 to 10.2) at each time point, respectively.  We are then left with the question whether this small difference reflects the underlying truth or chance.

Do the secondary aspects of these data validate the difference?  The expected advantage of 1:1:1 resuscitation is the warding off of evil spirits associated with transfusion-related coagulopathy – and we see in this study the primary driver of differences in mortality was related to deaths secondary to exsanguination.  Likewise, a greater number in the 1:1:1 group achieved satisfactory hemostasis.  So, using a Bayesian approach to interpreting the statistical tests for mortality, it is reasonable to adopt approaches based initial 1:1:1 resuscitation when massive transfusion is necessary, despite the limitations of the evidence.

One oddity worth noting in these data were the relatively small differences in sepsis and ARDS in the 1:1:1 group.  Increased use of FFP and, in particular, platelets are associated with these transfusion-related complications – and it has always been of particular interest whether a 1:1:1 ratio is safe, for precisely these reasons.  The inclusion of platelets in the 1:1:1 randomization may also be a matter for debate; few patients had any indication for platelets following the intervention, and further work could consider the relative utility of aggressive use of platelets.

Overall, however, this is best evidence to date the 1:1:1 ratio is a worthy initial target.

“Transfusion of Plasma, Platelets, and Red Blood Cells in a 1:1:1 vs a 1:1:2 Ratio and Mortality in Patients With Severe Trauma”

Friday, February 6, 2015

Which Review of Tamiflu Data Do You Believe?

Ever since its introduction, there have been skeptics regarding the utility of oseltamivir and other neuraminidase inhibitors for the treatment of influenza.  Roche has profited tremendously off strategic stockpiling by many governments as a response to pandemic influenza – yet, nearly all the data comes from Roche-conducted trials, and the data has been persistently cloaked from independent review.  This past year, after much strife and public shaming, the Cochrane Collaboration received some access to clinical trial reports to conduct an independent review.  This review found, on average, adults receiving early treatment with oseltamivir benefited by reduction in symptom duration from 7 days to 6.3 days.  No benefit was found for reduction in respiratory infectious complications or hospitalization, the truly critical need during influenza outbreaks.

However, a second group also conducted an independent review – the “Multiparty Group for Advice on Science”.  Their results, based on an individual-patient meta-analysis, are published in the Lancet and offer similar – yet wildly different – conclusions.  They find, as did the Cochrane group, approximately a 17-hour reduction in symptoms in the intention-to-treat population across the eight Roche trials evaluated.

Similar to the Cochrane review, they perform secondary analyses for “lower respiratory tract infection”(e.g., bronchitis or pneumonia) and hospitalization, stratified by ITT and ITT-infected populations.  Most prominently emphasized are the results for the ITT-infected population, in which the antibiotics for LRTI were provided to 4.2% in the oseltamivir cohort, compared with 8.7% in placebo.  Likewise, 0.9% of patients were hospitalized for any cause compared with 1.7% of placebo.  The authors therefore conclude oseltamivir use decreased infectious complications of influenza.

These numbers, however, are entirely different from the Cochrane review.  The Cochrane review found a 1.4% hospital admission rate in the oseltamivir cohort and 1.8% in the placebo cohort.  Broken down by trial, the admit rates for the oseltamivir cohort in the MUGAS analysis compared with the Cochrane review:
  • M76001: 7/965 vs. 9/965
  • WV15670: 1/241 vs. 1/484
  • WV15671: 1/210 vs. 6/411
  • WV15707: 2/17 vs. 2/17
  • WV15812+: 6/199 vs. 9/199
  • WV15819+: 6/360 vs. 9/362
  • WV16277: 2/226 vs. 2/225
The differences in WV15670 and WV15671 appear to stem, at least in part, due to the MUGAS analysis being restricted to only trial patients taking 75mg twice daily, and not 150mg twice daily.  However, it is otherwise entirely unclear how the Cochrane group found extra hospitalizations in the other trials the MUGAS group did not – particularly considering the hospitalization numbers in the placebo cohorts were essentially identical.  Might it be partly a result of the MUGAS group receiving their data directly from a Roche web portal, while the Cochrane group reviewed the individual clinical study reports?

Rather, might it be revealing to pry into the genesis of the “Multiparty Group for Advice on Science”?  Is it an unbiased, independent clearinghouse for re-analysis of trial data?  Do they have a long track record of respected publications in multiple disciplines?  Unfortunately, neither of these conjectures are true – making it increasingly likely they are a puppet foundation fraught with conflict-of-interest.  MUGAS and the present work were funded by an unrestricted grant from Roche.  Furthermore, MUGAS, along with the European Scientific Working group on Influenza (ESWI), are projects of Semiotics, a scientific branding and communication company specializing in influenza.  The stated goal of Semiotics is promoting corporate science and ensuring its place on top of the policy agenda – and MUGAS is one of their “brands”.  This ought to very clearly demonstrate MUGAS is not a scientific enterprise, and rather an organization tasked with the sort of advocacy as best represents the needs of its sponsors.

Any bias might also be clear just in the style used to present results.  These authors present the tiny absolute differences in hospitalization and infectious complications in forest plot figures using only relative risk, rather than absolute risk.  This serves to inflate the apparent effect size.  Conversely, they present the increased incidence of adverse effects in a table culminating in adjusted absolute risk, with the opposite effect.  This manner of presentation persists in their Discussion, highlighting a "significant 63% reduction in risk of hospitalization", compared with "absolute increases of 3.7% for nausea and 4.7% for vomiting."

So – the results of an analysis performed by a “brand”, highlighting results discordant with a prior unbiased analysis.  Where is the peer review vetting such discrepancies?  With so many professional reputations and so much revenue at stake – which report do you believe?

“Oseltamivir treatment for influenza in adults: a meta-analysis of randomised controlled trials”

Additional editorial content:
"The BMJ Today: The FDA and CDC’s disagreement over Tamiflu, and the spy who isn’t"