Admission risk markers for upper gastrointestinal bleeding: Can urgent endoscopy be avoided?

It is Friday night, and you are the on-call gastroenterologist at a busy regional hospital. You are called about a 55-year-old woman presenting with melena. There are no signs of shock, the hemoglobin and blood urea are normal, and the rectal examination reveals, yes, melena. Could you say: ‘send her home and ask her to come to the endoscopy unit fasted Monday morning?’ What are the possible consequences of this decision?

Upper gastrointestinal bleeding (UGIB) lends itself to systems-based thinking. There is a wide range of clinical outcomes, which can be reliably predicted according to long-established clinical and endoscopic signs. Putting these signs into a numerical score is attractive, as it may help to stratify management according to risk. For example, those at high risk of re-bleeding or death can be triaged to urgent endoscopy and intensive monitoring. More recently, the reverse argument has been applied: perhaps those at low risk of re-bleeding or death may be discharged—even without emergent endoscopy.

The most well-known clinical scoring systems for UGIB are the Baylor Bleeding Score,¹ the Cedars-Sinai Medical Center Predictive Index,² the Rockall Score,³ and the Blatchford Score.⁴ Of these, only the Blatchford Score can be derived without use of endoscopic findings. The Blatchford Score was developed and validated in a hospital population in West Scotland and was correlated against the Rockall Score. It was designed to establish which patients needed intervention: blood transfusion, surgery, endoscopic intervention to control bleeding, or if they had undergone no intervention but had died, re-bled, or had a substantial fall in hemoglobin concentration after admission. The latter are all features one would wish to avoid in a patient sent home without endoscopy.

There have been reports of outpatient management of UGIB;^5,6 however, in these studies all patients received urgent endoscopy as part of the work-up. But while larger centers, particularly those with gastroenterology trainees, invariably provide after-hours endoscopy on-call,⁷ this is not the case at many non-urban and satellite centers. These hospitals may not have ready access to out-of-hours endoscopy, and yet may have easy access to day-only facilities. A patient presenting with low-risk UGIB might be assessed and given appropriate therapy (such as cessation of non-steroidal anti-inflammatory drugs and commencement of proton pump inhibitor) with a view to outpatient endoscopy within, say, 1 week. Unless this practice can be bolstered with good data, most physicians would be reluctant to take the risk. Is availability of these ‘good data’ a reasonable prospect?

Now, two articles address the above scenario. In this issue of Journal of Gastroenterology and Hepatology, Masaoka et al. report a retrospective analysis of their hospital population.⁸ They performed record review and applied the Blatchford scoring system (see Table 1) to patients who presented with signs of UGIB and underwent emergency endoscopy in the Emergency Department of their hospital in 2004—1.3% of all hospital admissions for the period. They found that 70 of 93 patients (75%) required blood transfusion, operative or endoscopic interventions for the control of hemorrhage. A Blatchford score of <2 was determined as a cut-off value to classify the ‘high-risk’ group (needing the above interventions) from the ‘low-risk’ group (not needing the above interventions). In their small cohort, this cut-off was 100% sensitive (all those needing intervention were correctly classified) but only 13% specific. Indeed, only three of the 23 ‘low-risk’ patients were so classified. Using this strategy in the authors' population would have resulted in three discharges per year, compared with 90 admissions and urgent endoscopies—hardly a major cost-saving.

In another recent study, Romagnuolo et al.⁹ examined the Canadian RUGBE dataset of 1860 patients who underwent endoscopy for non-variceal UGIB. The 30-day re-bleeding rate was 18% and 5.3% died. If the modified Blatchford score (mBRS; modified to account for absence of urea level in the RUGBE database) was <2, the rate of re-bleeding was 5% and death occurred in 0.5%; 9.8% of patients had an mBRS of ≤1. Multivariate analysis showed that the mBRS < 2 was a significant predictor of better outcome (odds ratio 0.24 for re-bleeding and 0.12 for death). The authors commented: ‘A prospective, preferably randomized, study is required to validate whether this selected group of patients, especially those presenting after hours or on weekends, can be safely managed with neither emergent endoscopy nor expectant inpatient monitoring  . . .’

The problem with studying re-bleeding and death in UGIB is that these ‘serious mistakes’ are low-frequency events (0.5% rate of death in the low-risk group of the Romagnuolo study, and remember that these were patients that were admitted to hospital and had endoscopy). To conduct a prospective study that determined the difference between a ‘mistake’ rate of 0% (as suggested by the data of Masaoka et al.) and an arbitrary 1% (1 in 100 patients deemed to be low-risk readmitted with further bleeding), would be prohibitive in terms of sample size. For centers not able to perform urgent endoscopy, the current choice is between admission to hospital followed by ‘acute, non-urgent’ endoscopy, and discharge from hospital followed by ‘early’ endoscopy. For the sake of, at most, a 10% reduction in admission rate, why take the risk of a death rate this large?

The British Society of Gastroenterology guidelines for UGIB¹⁰ recommend as follows: ‘Patients with upper gastrointestinal bleeding should be admitted or transferred to a safe environment . . . endoscopy is undertaken . . . as a semi elective procedure in patients who have had relatively minor bleeding’. These guidelines are silent on whether patients may be safely discharged after a period of stabilization, before endoscopy. Even in high-risk patients, most re-bleeding occurs within 48 h,^11,12 suggesting that discharge after endoscopy and at least 48 h of observation could be appropriate.

So what do you advise, on that dark Friday night? Without clear evidence of safety, it would be unwise to depart from these published guidelines.