Upper GI Bleeding, “Answers”

1. Do you use a PPI infusion in patients with undifferentiated UGIB?

Acute upper gastrointestinal hemorrhage (UGIB) is a potentially life threatening condition caused by a number of etiologies. It is defined as any lesion causing bleeding proximal to the ligament of Trietz. This includes Mallory-Weiss tears, Boerhaave’s syndrome, esophageal varices and arteriovenous malformations. The most common cause of UGIB, however, is peptic ulcer disease (Lau, 2007). Proton-pump Inhibitors (PPIs) were first investigated for use in patients with peptic ulcer disease (PUD). PPIs act by decreasing the production and secretion of gastric acid. They irreversibly block the hydrogen/potassium ATPase in gastric parietal cells. Gastric acid has been shown in in vitro experiments to impair clot formation, promote fibrinolysis and impair platelet aggregation (Chaimoff, 1978; Green, 1978). Thus in theory, inhibition of gastric acid would allow the pH to rise, promoting clot stability and decreasing the likelihood of rebleeding. However, this benefit is only theoretical. The goal of raising the gastric pH above 6 has not been shown to be a reliable proxy for treatment efficacy (Gralnek, 2008).

EML UGIB AnswersIntravenous PPIs have become standard care post-endoscopy and post-operatively to prevent rebleeding in patients with PUD (Gralnek, 2008). Recent meta-analyses show that PPIs decrease the rate of rebleeding and surgical intervention in patients with PUD after endoscopic intervention. A pooled analysis of 16 randomized controlled trials found that a bolus dose of PPI followed by an infusion is more effective than bolus dosing alone for reducing the rebleed rate and the need for surgery, leading to the current recommendation (Morgan, 2002). However, this study clearly states: “Intravenous proton pump inhibitors appear to be useful in the prevention of rebleeding in patients with acute peptic ulcer bleeding that has been successfully treated with endoscopic hemostasis.”

A meta-analysis of 24 randomized trials (4373 patients) from the Cochrane group reached similar conclusions (Leontiadis, 2004). In patients with PUD, PPI treatment reduced rebleeding (NNT = 15), surgical intervention (NNT = 32) and repeat endoscopy (NNT = 10). However, they found no change in mortality (OR = 1.01). Overall, outcomes were modest. PPIs prevented rebleeds in 6.6% of patients, surgical interventions in 3.2% of patients and repeat endoscopy in 10% of patients. Interestingly, the Cochrane group did separate analyses of Western and Asian populations. They found that trials conducted in Asia demonstrated benefits to PPI infusions in peptic ulcer disease in terms of mortality (NNT = 34), rebleeding (NNT = 6) and surgical intervention (NNT = 23). Conversely, Western patients showed a suggestion of increased harm in PPI groups, although it was not statistically significant).

So, the current literature suggests PPI infusions in patients with known PUD offer only marginal benefits overall and possible harm in certain populations. What about in undifferentiated UGIB? Are PPIs beneficial in undifferentiated UGIB? Are they beneficial when given prior to endoscopy?

Fortunately, a Cochrane Review published in 2010 helps to address these questions (Sreedharan, 2010). This group found six randomized control trials (RCTs) relevant to the question. Of these, four compared PPI to either placebo (Daneshmend, 1992; Hawkey, 2001; Lau, 2007) or no treatment (Naumovski, 2005) and the other 2 studies compared PPI to H2 blockers. The studies comparing PPI to placebo (n = 1983) are the most relevant to our question.

The Cochrane review found no difference in mortality comparing PPI to placebo (OR = 1.19 (0.75-1.68)), no difference in rebleeding within 30 days (OR = 0.87 (95% CI 0.66-1.16)) or surgery within 30 days (OR 0.90 (0.64-1.27)). One of the limitations of this review was that PPI treatment was not the same in all studies. Only the Lau study compared a PPI bolus and infusion with placebo. In this trial, the authors reported that fewer patients in the PPI arm required intervention during endoscopy. However, there were no differences in any patient oriented outcome (death, rebleeding, or surgery within 30 days).

Bottom Line:

–PPI treatment in undifferentiated UGIB does not appear to decrease any clinically important effects including rebleeding, need for surgery, or death.

–PPI treatment prior to endoscopy for undifferentiated UGIB decreases the number of patients who require an endoscopic therapy during endoscopy.

2. Do you use octreotide in patients with bleeding varices?

There is little in the world of Emergency Medicine that gets a clinician’s pulse racing as much as the massive upper GI bleeder who has a history of esophageal or gastric varices. These patients have a high morbidity and mortality rate even with aggressive ED management, transfusion, intensive care, and gastroenterology consultation. Six week mortality is estimated between 11-20% (Dell’Era, 2008). Patients often are aware that they have varices, which aids in guiding treatment. However, in patients with no prior diagnosis or those too sick to communicate, establishing the presence of varices may be more difficult.

Octreotide is a somatostatin analog that acts by decreasing blood flow into the portal circulation, thus decreasing portal pressure, particularly postprandial flow (Abraldes, 2002). It is widely used for the treatment of variceal bleeding. In clinical trials, octreotide has only been noted to be beneficial when paired with endoscopic therapy (Dell’Era, 2008). A meta-analysis in 2001 demonstrated an improved efficacy of endoscopic therapy in terms of early rebleeding when octreotide was given concomitantly (Corley, 2001). The study found an NNT of 6 when compared to placebo for rebleeding and transfusion requirement. The reduction in transfusion was modest (about 0.7 units of PRBCs). Additionally, no study found any reduction in mortality or overall rebleeding (only benefit in early rebleeding) (Corley, 2001; Dell’Era, 2008; Longacre, 2006). A more recent randomized, placebo controlled study found that sclerotherapy plus octreotide was equal to sclerotherapy plus placebo in terms of 7-day mortality, rebleeding, transfusion requirements, and ICU stay (Morales, 2007).

In 2008, the Cochrane group performed a systematic review of all randomized trials looking at octreotide in the treatment of varices (Gøtzsche, 2008). The group found 21 randomized trials of octreotide versus placebo (or no treatment). They concluded that the use of octreotide did not reduce mortality. It did reduce the amount transfused by about ½ a unit of blood in the studies with a low risk of bias (1.5 units in those with high risk of bias) but this result was not thought to be clinically significant by a number of commentators. They found no difference in rebleed rates in the low-risk of bias studies (but a substantial reduction in the high risk of bias studies). The Cochrane group did find a lower rate of failed initial hemostasis in the octreotide treatment group.

Overall, it seems that the incremental benefit of octreotide in addition to endoscopic therapy in the treatment of variceal bleeding is only seen in surrogate, non-patient-oriented outcomes. No study, systematic review, or meta-analysis has shown a benefit in mortality. Additionally, this is for patients in whom it is known that varices are the cause of their UGIB. There is no data on the use of octreotide in undifferentiated UGIB . Additionally, there is no pathophysiologic basis for its use in these patients.

Bottom Line:

–Octreotide in combination with endoscopic therapy in patients with bleeding varices has not been shown to reduce the rate of failed hemostasis, rebleed rate, or mortality.

–Octreotide did show a modest reduction in blood transfusions required in this clinical scenario.

3. In which patients with UGIB do you place a nasogastric tube (NGT) and for what purpose (diagnostic vs. therapeutic)?

Historically, nasogastric intubation (NGI) in patients with suspected upper gastrointestinal bleed (UGIB), has served multiple roles: therapeutic, diagnostic, and prognostic. However, its utility has been controversial for years. Recognizing the severe discomfort of this basic procedure and its rare but potential complications, scrutiny of the extant data is warranted to decide what, if any, benefit NGI provides.

In a patient with GI bleeding, distinction between upper and lower source (i.e., bleeding proximal or distal to the ligament of Treitz), is essential for determining advanced management. Gross evaluation of nasogastric aspirate (NGA) is not an uncommon diagnostic test done in the ED to assess for the presence of a proximal bleed. Hematemesis is virtually always the result of briskly bleeding lesions proximal to the pylorus (Peura, 1997), making NGA obsolete in the diagnostic evaluation of patients with hematemesis. In patients with melena or hematochezia without hematemesis, the source is not as straightforward and NGA has been recommended for diagnosis. It stands to reason, then, that studies to determine the diagnostic utility of NGA should focus on patients without hematemesis. In one such study, the yield of positive (i.e., bloody) NGA was significantly lower than in prior studies that included subjects with and without hematemesis (Witting, 2004). This likely reflects that UGIB without hematemesis often results from either a slower bleed or one distal to the pylorus, such as a duodenal lesions. Both of those scenarios are much less likely to yield a positive result with NGI. A technically adequate NGI should include duodenal aspirate, as evidenced by presence of bile, but often do not. Although a positive NGA was determined to strongly predict an UGI lesion on endoscopy, with a likelihood ratio (LR) of +11 and a positive predictive value (PPV) of 92%, a negative NGA showed minimal utility with a LR of 0.6 and NPV of 64% (Witting, 2004).

Other diagnostic models have been shown to have utility without incorporation of NGI. Three strong independent risk factors (age<50, black stool, and BUN/creatinine ratio greater or equal to 30) were identified to predict an UGIB in patients without hematemesis and compared favorably to NGA. The absence of all three risk factors corresponds to a 5% risk of UGIB. Of those with 2 or more risk factors, 93% had an UGIB (Witting, 2006).

Historically, NGI has also contributed to assessing prognosis, helping to risk stratify patients, guide type and timing of intervention, as well as influence disposition. The prognostic value of NGI has been explored in several studies. A positive NGA has been associated with higher mortality rates (Leung, 2004) and worse outcomes (Stollman, 1997). Coffee-ground or bloody NGA has been shown to strongly predict the presence of high-risk endoscopic lesions (HRL), defined as an active bleed or visible vessel (Perng, 1994). Bloody NGA demonstrated an increased association with HRL when compared with clear or bilious NGA and with coffee-ground NGA, having an odds ratio (OR) of 4.8 and 2.8, respectively (Aljebreen, 2004). For a test to be clinically useful, however, it must be sensitive enough to rule out the dangerous condition. If “positive NGA” refers only to gross blood, the PPV and NPV of HRL are 45% and 78%, respectively; if defining “positive NGA” to include all aspirate except clear or bilious, the PPV and NPV are 32% and 85%, respectively (Aljebreen, 2004). Therefore, even using the most inclusive definition of a positive NGA, this technique misses a significant portion of patients with high risk lesions. If NGA is grossly positive, then such a lesion is likely. However, based on this data, a negative result has a poor negative likelihood ratio, and lacks the sensitivity to rule out a HRL.

The ability to identify patients with HRL, lesions likely to re-bleed, and those with higher mortality is important. These correlations, however, do not intrinsically translate to useful information in terms of acute management decisions. A more clinically useful endpoint would differentiate those who would benefit from urgent endoscopy from those who may safely wait. Noninvasive prognostic scales, such as the Glasgow-Blatchford and pre-endoscopic Rockall scores, have been developed and validated to accurately predict patients who require intervention and who can be safely wait (Barkun, 2010). The Glasgow-Blatchford score has been determined to have sensitivity approaching 100%, allowing the provider to rule out significant bleeding (Chen, 2007; Srygley, 2012).

No one test or scoring system has yet been devised to stratify those who require emergent versus urgent endoscopy. Likewise, the benefit of rapid endoscopy <24 hours is not clear (Targownik, 2007; Spiegel, 2009). A 2010 prospective study showed mortality benefit from endoscopy less than thirteen hours after presentation in high-risk patients (based on the Glasgow-Blatchford score) (Chin, 2011). Although a frankly positive NGA highly predicts a HRL, the clinical picture of the patient is more likely to determine the rapidity with which endoscopy should be performed.

In terms of therapeutic use, the importance of adequate gastric lavage was highlighted in a retrospective study revealing increased morbidity and mortality associated with inability to clear fundal blood during endoscopy (Stollman, 1997).These results postulate that inadequate pre-endoscopy gastric lavage may be to blame for poor visualization and thus worsen outcomes. No data exists to examine the impact of gastric lavage on the “fundal pool.” However, multiple adjunctive measures have been shown to be effective in clearance of blood such as endoscopic lavage or use of promotility agents such as metoclopramide or erythromycin (Leung, 2004). International consensus guidelines recommend the use of promotility agents only in the select group of patients where a significant amount of blood is anticipated (Barkun, 2010).

Looking at overall impact of NGI on patients, Huang, et al. observed the outcomes of clinically matched patients with UGIB, comparing those who had NGI and those who did not. Although patients with NGI obtained endoscopy more quickly on average than those without NGI, no significant difference in length of hospital stay or 30 day mortality was found (Huang, 2011).

The diagnostic utility of NGI is limited. It adds no useful data in patients with hematemesis and is not sensitive in patients without. The ability of NGI to predict patients with high-risk lesions is good but, again, lacks the sensitivity to rule them out, proving inferior to some noninvasive prediction scores. NGA alone is not adequate to risk-stratify patients and change the urgency of obtaining endoscopy. The therapeutic value of NGI for pre-endoscopic gastric lavage is unclear and reasonable alternatives exist which spare patient discomfort. NGI has not been shown to improve patient outcomes.

Bottom Line:

–A negative NGI cannot rule out UGIB or the presence of high risk lesions

–The color of NGA is inadequate data on which to base management decisions

–NGI does not clearly improve endoscopic results or patient outcomes

4. What is the utility of fecal occult blood test for patients in whom we suspect UGIB?

Fecal occult blood test (FOBT) is intended as a screening tool for lower GI malignancy but is commonly used in the Emergency Department as part of a work-up for suspected UGIB. Virtually all literature surrounding the test relates to its use in the outpatient setting. Without dedicated evidence to support the use of this test in the ED, we should take efforts to be knowledgeable about the test itself, and thoughtful regarding when to use it and how to interpret it.

Different types of FOBT are available and knowing the exact test used is important for proper interpretation. Non-guaiac based FOBT have increased specificity to LGIB as they use immunochemical assays to detect human hemoglobin, not heme. They are not useful in detection of UGIB, as most hemoglobin is digested in the small intestine and not present in rectal stool (Allison, 2007). Guaiac-based tests (gFOBT) detect heme by using it as a catalyst in the oxidizing reaction of the guaiac-impregnated card producing an immediate blue color where heme is present. The result can be altered by a variety of substances, which participate in this reaction. Heme-containing red meat or peroxidase-containing foods (turnips, radishes) can produce a false positive. Vitamin C may produce a false negative due to its antioxidant effect. The sensitivity varies depending on the specific test used. It increases with the amount of blood present. The Hemoccult II (guaiac based test) requires 10ml of fecal blood loss per day (10mg blood/gram of stool) for 50% sensitivity, but may be positive with <1mg/g (Stroehlein, 1976). In contrast, melena production requires 50mL gastric blood, based on a study by Schiff, et al. in 1942. Despite the apparent simplicity of the gFOBT, it is user-dependent. In a survey of 173 medical providers, 12% did not accurately interpret results (Selinger, 2003).

To determine whether FOBT enhances our management of a patient with suspected UGIB, certain questions should be answered. Do the results of FOBT have diagnostic value in these patients? Is FOBT sensitive enough that a negative test can rule out UGIB? Does the false positive rate lead to significant undue intervention such that the risks of the test outweigh the benefits?

Stool color is among the best clinical predictors of UGIB. According to a literature review, if a patient reports melena, the likelihood of an UGIB is increased more than five-fold; if found on exam, UGIB is 25 times more likely to exist (Srygley, 2012). Black stools were shown to be 80% sensitive and 84% specific for an UGIB (Witting, 2006). Conversely, blood clots present in stool make UGIB 20 times less likely (Srygley, 2012). Given opposing correlation to UGIB based on the color of stool, both of which would be theoretically guaiac-positive, the role of gFOBT in either black or red stool would only be to distinguish blood clot from red food particle. The significance of guaiac-positive brown stool in a patient with history concerning for UGIB, however, is not evident. Similarly enigmatic is the significance of guaiac-negative stool.

One to two liters of ingested blood may cause melena for up to five days, starting approximately four to 20 hours after its ingestion (Wilson, 1990). It can be inferred that guaiac negative stool may occur in active UGIB if the blood-containing stool has not had sufficient time to reach the rectum, or if the bleeding has been intermittent and the sample obtained represents a non-bleeding interval. Although it would be difficult to imagine a significant bleed occurring for >24 hours not resulting in a positive result, it cannot rule out the possibility.

The false positive rate of gFOBT in predicting acute UGIB is not known. A review article looked at the utility of endoscopy to detect upper GI lesions in non-emergency patients with positive screening FOBT and a negative colonoscopy. Of patients with guaiac-positive screen, 37-53% have negative colonoscopies. Of these patients, the literature review showed endoscopy to be positive for UGI cancer in <1%, positive for nonmalignant sources of bleeding 11-21%, and incidental, likely unrelated, findings in 10-36%. The review did not extrapolate data to differentiate results of patients with anemia or other significant symptomatology, which may have been interesting for purposes of this discussion. Although this data does not apply to our patient demographic, it does give some insight into the low specificity and low PPV (for this population) of guaiac positive, non-melenic stool in determining endoscopic pathology (Allard, 2010).

As with any diagnostic test, in order for it to have utility, it must have a reasonable sensitivity and specificity to avoid both missed diagnosis and excessive overtreatment. While the sensitivity of gFOBT to detect blood is high, its ability to detect UGIB is unknown. Therefore, it cannot be used to rule out UGIB if a high clinical suspicion exists. The extremely low specificity poses the dilemma of what to do with guaiac-positive brown stool. Without a sufficient amount of blood to produce melenic stool, can a positive guaiac test be discounted as clinically insignificant in the ED or does it commit the provider to pursuing medical and endoscopic management? Undoubtedly, the results of gFOBT alone should not dictate care, but the question remains as to whether or not occult blood testing should be obtained at all in ED evaluation of UGIB. Unfortunately, using this test in a setting for which it was neither intended nor researched limits our ability to interpret its results, imparting the risks associated with misinterpretation.

Bottom Line

–The role of gFOBT in evaluating acute UGIB has not been sufficiently studied

–Stool color (black or red) has more diagnostic value than gFOBT results

–Positive gFOBT does not rule in UGIB and carries the potential risk of unnecessary treatment or procedures

–Negative gFOBT does not rule out UGIB and risks a false sense of security and under-treatment of true disease

This entry was posted in Uncategorized and tagged , , , , , , . Bookmark the permalink.

7 Responses to Upper GI Bleeding, “Answers”

  1. Great review, thanks!

    If I may suggest, I think the only part missing is a discussion about the off-label use of tranexamic acid… I’ve used it in a few cases of brisk bleeders, after the CRASH data showed no harm for a short infusion… Hope some data comes out you can can crunch for us!

    cheers!

    • Great point. I recommend using TXA as well but we try to stick with the lit for EM Lyceum.

      • Thanks for the great review. Hopefully the HALT-IT trial will soon be able to give us the evidence for TXA in Upper GI bleeding. Over 1000 patients recruited so far: http://haltit.lshtm.ac.uk

        One other question. What Hb are you targeting? Here was our take on the Villanueva trial: http://www.wessexics.com/The_Bottom_Line/Review/index.php?id=8999282330351501733
        Would be interested to know your thoughts

        Thanks

      • emlyceum says:

        Thanks for reading, and for pointing out the HALT-IT trial: With their goal of 8,000 patients (whoa!), it looks like it might be 2016 or 2017 until we have any results. Definitely worth watching, and maybe they’ll publish an interim analysis.

        Anand may want to weigh in as well, but I typically do use the restrictive strategy in the Villanueva study as my target (7-9), in patients similar to those included/excluded by their criteria (not minor or massive, no recent vascular issue, etc.). One issue I’ve encountered, which you aptly point out, is failing to achieve early endoscopic intervention.

        -Whitney Bryant, for emlyceum (@emlyceum)

  2. Matt says:

    For the severe UGIB with hematemesis who’s going for emergent endoscopy I still think that there’s utility in placing an NGT. In case things turn sour and you have to intubate them before they go to the ICU/GI suite it’s reassuring to know that there isn’t a time bomb of blood possibly waiting for you upon induction. It’s also my routine to intubate those patients in reverse Trendelenburg, still obeying the external auditory meatus to sternal notch.

    Some might say that placing the NGT through the esophageal sphincter into the stomach “stents” it open and provides and easier egress for any residual content to aspirate, but I don’t think that outweighs the potential benefit of decompressing that stomach.

    • Good point, Matt. If they are actively hematemisizing (not sure that’s a word) I intubate prior to scope so as not to run into more problems. Can always extubate after the procedure. I think it’s reasonable to drop an NGT for therapeutic purposes for active hematemesis but diagnostically, NGT has limited utility.

  3. Pingback: The LITFL Review 134 - LITFL

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s