When confronted with a patient who exhibits evidence of a recent upper GI bleed that is now appears to have stopped bleeding, what does the best evidence say about any real benefit to giving a proton pump inhibitor (Protonix, Prevacid, Nexium, etc) over an H2 Receptor Antagonist (Pepcid, Zantac, Axid, etc)?
It doesn’t improve mortality, except in select groups. But it might lessen the chance of surgery or re-bleeding. If, however, they are actively bleeding, their mortality may be improved.
In the ED, upper gastroenterological (UGI) bleeding is a common presentation with a 10-20% mortality essentially unchanged over the last 50 years. UGI bleeds in the US affect 100 people per 100,000 population resulting in 250,000 annual admissions and $1 billion per year. While Proton Pump Inhibitors (PPI) have been available for over a decade, individual studies and past meta-analyses have failed to demonstrate an impact on mortality when used for this indication. The Upper GI Cochrane group updated a 2004 Cochrane Review in 2006 re-addressing this question. One of the many benefits of the Cochrane Collaboration is that they are updated every 2-4 years for most topics to determine whether evolving evidence changes the previous results.
The updated PPI in UGI bleed Cochrane Reviews contains some interesting findings. PPI’s do not improve (or worsen) mortality, but they do reduce re-bleeding rates (Number Needed to Treat or NNT = 13, 95% CI 10-25), surgical intervention (NNT = 34, 20-50), and requirement for further endoscopic treatment after randomization (NNT = 10, 8-17). These findings were maintained even when the SR authors removed individual trials or analyzed only the most methodologically sound trials. Asian trials, however, did demonstrate a significant mortality reduction. Similarly, trials which included only patients with active bleeding or a non-bleeding visible vessel also showed significantly reduced mortality. Finally, those patients with high-risk endoscopic signs with high-dose (80 mg omeprazole IV followed by 8mg/hour IV infusion for 72 hours) PPI treatment also showed a mortality benefit. No PPI was found to be superior to another PPI in this meta-analysis. Note that the comparison group for PPI was either a histamine Type 2 receptor antagonist (H2 RA like famotidine) or placebo. The authors do not address PPI Number Needed to Harm, but the relatively mild side-effect profile of these agents would suggest limited risk in the setting of an ongoing UGI bleed.
When confronted with an actively bleeding peptic ulcer in your ED, PPI’s are superior to H2RA to minimize re-bleeding, surgical intervention, and the need for repeat endoscopy. A physician would have to treat 13 such patients with PPI’s to prevent one from re-bleeding. Although UGI-related deaths are not reduced by PPI use in this setting, certain subsets of patients (Asians, endoscopic arterial “pumper,” or visible non-bleeding vessel) do demonstrate mortality benefit from PPI and high dose IV PPI may be more beneficial than oral PPI in these higher risk patients.
What is the “best evidence”? What is the Cochrane Review and why do you rely on their analysis?
You don’t rely on one review, just as you wouldn’t rely on one study. But you take into account a good comparison of the relevant research. And the Cochrane Review is one such comparison.
Physicians today are bombarded with information – up to 5000 bio-medical articles are published each day! No surprise that many of us feel overwhelmed in trying to maintain cutting-edge proficiency.
While the first precept of EBM is that evidence alone is never enough (but rather evidence, clinical expertise, and patient preferences are equally important), the second principle is that not all evidence is created equally, leading to the concept of a hierarchy of evidence. Picking up a random medical journal, any physician can quickly come to the same conclusion. Some studies are simple case reports of a single observation by physicians, without attempting to control for confounding influences. Other articles are simple Narrative Reviews (NR) where authors summarize a topic via their analysis of the literature they found and decided to include. These differ from Systematic Reviews (SR) in that narrative reviews fail to describe an explicit, reproducible search strategy and never “grade” the included evidence for quality. Quality in this case means how well individual studies guard against bias (a systematic difference that alters the results). Many forms of bias exist in research, but suffice to say that one of the best ways to avoid bias is the randomized, blinded clinical trial in which researchers, patients, and outcome assessors do not realize which intervention a given subject received. Unfortunately, not every clinical question can be ethically evaluated by a Randomized Controlled Trial. One probably wouldn’t want to volunteer for a study to test the lifelong effect of smoking on cancer risk and agree to smoke cigarettes or placebo-cigarettes for the next 40 years!
Again, all evidence is not created equally and SR are no exception. Systematic Reviews and meta-analyses must rigorously locate and appraise evidence relevant to a focused clinical question. Critics of SR appropriately point out that “garbage in is garbage out” meaning that if poor quality studies subject to bias are combined into a SR their results will mean no more to the reader than their individual studies. Before combining individual study results numerically into meta-analysis, SR authors must look at each study asking questions like “Were the patients, interventions, means of measuring outcomes, and outcomes measured similar enough to justify lumping these results together?” Statistically, SR authors do these study-to-study comparisons with tests of heterogeneity. Finally, useful SR’s will grade the quality of the trial evidence they include and weight “better” evidence as more important. Sounds like a lot of work, eh? Fortunately, a source for high quality, methodologically sound SR’s exist in the Cochrane Collaboration!
Glossary of Terms
Hierarchy of Evidence – the concept that certain study designs minimize bias better than others. With the concept of bias-minimization in mind, evidence can generally be rated in descending level of quality (see graph) Within Systematic Reviews of RCT, the highest quality SR are produced by the Cochrane Library.
Narrative Review – summary of broad-based question without specifying literature search strategy, study exclusion methods, or evidence appraisal with a non-quantitative summary.
Systematic Review – When done well, a systematic review will focus on a specific clinical question with an explicitly described (reproducible) search strategy, study exclusion methods, evidence appraisal, and a qualitative and quantitative assessment of study-to-study heterogeneity which might preclude lumping the findings together. The “subjects” of Systematic Reviews are the individual studies. When individual studies are sufficiently similar, their results can be mathematically combined into a Meta-Analysis. Every Meta-Analysis is a Systematic Review, but not every Systematic Review is a Meta-Analysis.
Number Needed to Treat – The number of subjects who need to receive an intervention in order for one additional patient to benefit.
Number Needed to Harm – The number of subjects who need to receive an intervention in order for one additional patient to have an adverse effect that wouldn’t have otherwise occurred.
Christopher Carpenter, MD is an Assistant Professor at Washington University School of Medicine and a speaker in the Best Evidence in Emergency Medicine lecture series.