Month: August 2014

Bizarrely Alarmist Pediatric URI Study

In our new Gawker and Buzzfeed-fueled, short-attention span reality, attention-grabbing headlines are essential.  So, let me come up with the modern headline for news coverage of this latest article, published in Pediatrics:  “Is your child’s next cold a killer?”

Seriously, as covered by Medscape (subscription required):

“As many as 1 in 3 children seeking treatment in the emergency department for influenza-like illnesses (ILI) at the peak of influenza season are at high risk of suffering severe complications, such as pneumonia.”

But, that’s hardly the case.  The study upon which they report is an observational cohort of ILI presenting to a tertiary children’s hospital.  To be eligible for inclusion, children needed to have ILI, defined as fever + cough/sore throat, and have “moderate to severe” symptoms.  However, their definition of “moderate to severe” is not based on any specific clinical criteria – it’s based off the surrogate of whether a clinician judged venipuncture and viral testing necessary.

So, 125,940 children were screened during the study period, and this cohort comprises the, presumably, sickest 241 of those.  Of those 241, over half had one of a predefined list of high-risk conditions: asthma, neurologic/neuromuscular disease, respiratory disease, heart disease, or immunosuppression.  And, yes, about 40% of each cohort developed a complication – most frequently pneumonia.  But, it should not be concluded there are killer viruses everywhere – rather, the sickest ILI, particularly those children who presumably appeared ill despite lacking underlying chronic illness, are the tiny cohort at higher risk of subsequent complication.

The authors also try to single out H1N1 influenza as an independent risk factor for subsequent complications.  11/29 patients with H1N1 influenza developed pneumonia, compared with 1/20 patients without, leading to their conclusion H1N1 confers particular risk.  However, 22/29 of patients diagnosed with H1N1 carried high-risk comorbidities, compared with only 10/20 in the non-H1N1 influenza cohort.  Yes, H1N1 probably increases risk of respiratory complications, but these data may not reliably support their conclusion.

“Severe Complications in Influenza-like Illnesses”
http://pediatrics.aappublications.org/content/early/2014/07/29/peds.2014-0505.abstract

A Moratorium on Steroids for TBI

A guest post by Rory Spiegel (@EMNerd_) who blogs on nihilism and the art of doing nothing at emnerd.com.

In 2004 the CRASH trial examining the efficacy of steroids for acute traumatic brain injury (TBI) was published in The Lancet.  This massive trial included over 10,000 patients was stopped prematurely because of an increased mortality in the patients who received corticosteroids. This should have definitively closed the book on such a therapy.  Despite this damning evidence, it appears all one has to do to make this question relevant again is to devise a disease-oriented endpoint with plausible clinical relevance and test it using a sample size too small to differentiate these harms from the surrounding noise of statistical chance.

Authors of the recently published Corti-TC trial did just this. Asehnoune et al examined the effect of the combination of hydrocortisone and fludrocortisone for the prevention of hopsital-acquired pneumonia (HAP) in patients with severe TBI. This is not the authors first foray into the efficacy of steroids for TBI. Their original trial was published in JAMA in 2011 and examined the effects of hydrocortisone to prevent HAP in patients having experienced poly-trauma. In this initial trial, about half the 149 patients randomized to either hydrocortisone or placebo suffered a severe TBI. The authors found that 35.6% of the patients in the active treatment arm developed HAP compared to 51.3% in the placebo group. This difference was seen exclusively in the subgroup of patients with TBI.  Thus the authors set out to validate these findings by solely examining patients suffering from acute TBI. As a harbinger of things to come, the authors justified the 3% increase in mortality as statistical chance, since it failed to reach statistical significance due to the small sample size.

In what essentially is a validation cohort the Corti-TC  trial was devised. Patients were randomized to either a 10 day course of both hydrocortisone and fludrocortisone or equivalent placebos. Cortisol levels were drawn before treatment was initiated, and in those found to be adrenally competent treatment was stopped. Once again the authors’ primary endpoint was the 28-day incidence of HAP as defined by a new infiltrate on chest x-ray with at least two of the following criteria; a temperature >38°C, leucocytosis >12 000 cells per mL, leucopenia <4000 cells per mL, or purulent pulmonary secretions.

Corti-TC demonstrated a similar difference in rates of HAP in patients given steroids vs those who received the placebo. Specifically 45% of the patients in the steroid group compared to 53% in the control group developed HAP over the first 28-days.  Although this difference did not reach statistical significance due to a lower than anticipated overall incidence of HAP, a relevant divergence between the active and control groups is evident. This difference remained consistent whether or not patients were found to have adrenal deficiency, indicating that cortisol levels do not predict a subset of patients who will benefit from steroids. Of concern is the 2% absolute increase in mortality of patients treated with steroids. This difference was observed primarily in the subset of patients later found to be adrenally intact. The authors once again justify this increased mortality by its failure to reach statistical significance (p-value of 0.32). That this exact trend was demonstrated in their original study goes unmentioned.  In fact, the same magnitude of harm caused the authors of the CRASH trial to halt their study prematurely. Given the collective consistency with which this mortality detriment has been demonstrated across trials it should not be written off as fluctuations of random chance. Interpreting this literature in its totality, it becomes obvious that these recent examinations of steroids in head trauma are vastly underpowered to detect the true harms involved with the utilization of such an intervention.

In the discussion section of both their trials, the authors question why their patients fared better than those in the CRASH cohort. They hypothesize that the overall higher acuity of their patients may be responsible for this difference in outcomes. The authors recommend further studies be performed to elucidate this uncertainty. I would argue that their cohorts fared no better than the CRASH patients. In fact, the absolute increase in mortality was identical to that of the CRASH trial. It is only because these authors defined success with a disease oriented outcome of little clinical significance(HAP), that their cohorts appear to fare better than the far more robustly powered CRASH cohort. At this point it seems clear steroids in acute TBI are harmful. Further studies to clarify the magnitude of benefit of irrelevant outcomes seem unwarranted.

“Hydrocortisone and fludrocortisone for prevention of hospital-acquired pneumonia in patients with severe traumatic brain injury (Corti-TC): a double-blind, multicentre phase 3, randomised placebo-controlled trial”
http://www.thelancet.com/journals/lanres/article/PIIS2213-2600(14)70144-4/fulltext

The tPA Cochrane Review Takes Us For Fools

It’s been 5 years since the last Cochrane Review synthesizing the evidence regarding tPA in acute ischemic stroke.  Clearly, given such a time span, in an area of active clinical controversy, a great deal of new, important, randomized evidence has been generated!

Or, sadly, the only new evidence available to inform practice is IST-3 – a study failing to demonstrate benefit, despite its pro-tPA flaws and biases.  So, it ought not be a very exciting update, considering the 2009 version included 26 trials, and the 2014 update now includes only 27 trials.  Their summary conclusion, with only additional evidence of regression to the mean, ought remain essentially the same, or even less optimistic, right?

Of course not:

2009:
Overall, thrombolytic therapy appears to result in a significant net reduction in the proportion of patients dead or dependent in activities of daily living. This overall benefit was apparent despite an increase both in deaths (evident at seven to 10 days and at final follow up) and in symptomatic intracranial haemorrhages. Further trials are needed to identify which patients are most likely to benefit from treatment and the environment in which thrombolysis may best be given in routine practice.

2014:
Thrombolytic therapy given up to six hours after stroke reduces the proportion of dead or dependent people. Those treated within the first three hours derive substantially more benefit than with later treatment. This overall benefit was apparent despite an increase in symptomatic intracranial haemorrhage, deaths at seven to 10 days, and deaths at final follow-up (except for trials testing rt-PA, which had no effect on death at final follow-up). Further trials are needed to identify the latest time window, whether people with mild stroke benefit from thrombolysis, to find ways of reducing symptomatic intracranial haemorrhage and deaths, and to identify the environment in which thrombolysis may best be given in routine practice. 

They added a neutral trial comprising 43% of the tPA subjects to the existing analysis, and now it can be decisively promoted “up to six hours”?  How is this conceivable?

So, in the most literal sense, technically, the authors’ statement is not untrue.  Analyses 1.12 and 1.13 aggregate all patients treated in trials between 0-6 hours, looking at mRS 0-2 and 0-1 at the end of follow up.  Indeed, for mRS 0-2, the OR favors thrombolysis at 1.17 (1.06 to 1.29), and for mRS 0-1, the OR favors thrombolysis at 1.29 (1.16 to 1.43).  Therefore, the authors are not falsely advertising tPA as beneficial for reducing death and dependency out to six hours – as long as you wear your pro-tPA blinders.

These authors, with multiple professional and financial conflicts-of-interest, simply choose to focus on inappropriate chunking of data for a theoretically time-dependent condition, rather than acknowledge their own analyses performed providing evidence to the contrary.  Analysis 1.21, in which they split out the patients treated into 0-3 and 3-6 hour cohorts, clearly demonstrates there is no basis upon which to claim benefit beyond 3 hours.  The OR for favorable outcome with thrombolysis in the 0-3 hour window is 1.53 (1.26 to 1.86), but the OR for 3-6 hours is 1.07 (0.96 to 1.20).  Then, the authors also neglect to mention Analysis 1.26, showing deaths are neutral between 0-3 hours, with an OR of 0.91 (0.73 to 1.13), but increased by thrombolysis in the 3-6 window, with an OR of 1.16 (1.00 to 1.35).

So, tPA after 3 hours: no functional outcome benefit and increased deaths – yet the authors are extolling the benefits of tPA to 6 hours?  There is no reasonable justification for such distorted reporting of their own analyses.  Simply unacceptable – and grossly misleading for the vast population of clinicians who do not or cannot access the full text, and only read the abstract.

Let’s be perfectly clear – I am not anti-tPA.  I am, however, opposed to the unfettered expansion of tPA as guideline-mandatory treatment to a larger eligible cohort – as is increasingly prevalent across contemporary literature, and fueled by manufactured-sponsored COI.  Acute ischemic stroke is a heterogenous disease, with varying underlying etiology and diverse cerebrovascular substrate.  It is clear there are subsets of patients for whom the likelihood of harm from tPA exceeds the benefit, and we ought to be using precision medicine to narrow the treatment population, not expand it.

Thanks to Rory Speigel (@EMNerd_) for alerting me to this publication.

“Thrombolysis for acute ischaemic stroke (Review)”
http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD000213.pub3/abstract

Nocebo Effects, the Dark Side of Placebo

We easily appreciate the placebo effect – the simple expectation of a treatment’s success positively affects its efficacy.  To prove a new treatment’s utility, then, we compare it against a placebo – a sham with the same expectation of success – to reveal a true magnitude of benefit (or harm).

However, much less appreciated is the flip side: nocebo effects.  I.e., if  patient expects to have adverse effects from a treatment, they are more likely to do so.  This has implications for clinical trials, of course, but also for discontinuation of therapy in general practice.  For example, consider those lovely pharmaceutical commercials, showing happy couples skydiving, in bathtubs, or otherwise living faux healthy lives – while simultaneously providing the droning voice-over detailing a litany of dire, disabling side effects.  Each mention of adverse outcome increases the likelihood a patient will perceive or experience it, and thereby potentially harm patients through decreased adherence to otherwise beneficial treatment.

Nocebo – Darren Cullen (2012)

These authors review the causes and implications of nocebo effects, and have several recommendations regarding effective strategies to minimize nocebo effects.  My favorite, by far:

“Refer to web-based and other information systems that provide evidence-based information, instead of unproven, anxiety-increasing comments.”

Ah, yes – you mean, basically, the entire Internet: insane, uninformed, anecdotal.  Good luck with that.

“Avoiding Nocebo Effects to Optimize Treatment Outcome”
http://www.ncbi.nlm.nih.gov/pubmed/25003609