Risk stratification scores

Well-validated prognostic scoring systems such as the Rockall et al15 or Glasgow-Blatchford Scale (GBS)16 scores should be used in all patients presenting with NVUGIB.8 Using combinations of clinical, laboratory and endoscopic data, these scoring systems facilitate risk stratification and identification of low-risk patients suitable for early discharge (either before or promptly after endoscopy), as well as those at higher risk for rebleeding, mortality and the need for therapeutic intervention at endoscopy. Despite their derivation over a decade ago, external validation in multiple cohorts and recommendations in evidence-based national and international guidelines,8 data from Canada and the UK show poor adherence to their use. In the Canadian RUGBE study, one out of 15 sites used the Rockall score routinely and explicitly to stratify patients with NVUGIB and none routinely used the GBS.17 Subsequently, in the REASON-2 study only one of 826 patients across 43 participating sites was found to have a Rockall or GBS score documented in the medical records.11 In the large UK national audit of UGIB and the Use of Blood, only 19% (1250/6750) of patients were found to have either a Rockall or GBS score documented in clinical charts.14 Although no randomised trial has directly compared clinical outcomes in patients who have versus those who have not been risk-scored, a wealth of observational data supports the benefits of the routine use of these scores, and they are a simple aspect of care which should be achievable in all patients admitted with UGIB and could be formally evaluated as components of future assessments.

Pharmacological treatment before endoscopy

There is evidence from randomised controlled trials (RCTs) and meta-analyses of RCTs of the net benefits associated with three main classes of drugs prior to endoscopy—proton pump inhibitors (PPIs), antifibrinolytics and prokinetic agents. A Cochrane meta-analysis of six RCTs including 2223 patients presenting with unselected UGIB comparing PPIs with control administrations (placebo or H₂ receptor antagonists) demonstrated significant reductions in the proportion of patients with high-risk stigmata (HRS) at endoscopy, as well as a reduction in the need for endoscopic therapy.18 Although it did not lead to reductions in the key endpoints of rebleeding, mortality or need for surgical intervention, downstaging of endoscopic stigmata has obvious benefits but has only been able to be demonstrated in cost-effectiveness modelling to date.19 Furthermore, the optimal dose and route of administration remain unclear. These considerations may explain why, in practice, there appears appropriately to be uncertainty among clinicians as to the usefulness of PPIs before endoscopy. In the UK audit, 43% of patients presenting with all-cause UGIB were prescribed a PPI before endoscopy,14 67% in the Italian PNED-1 registry were prescribed a PPI either during the emergency room visit or hospitalisation12 and, in the RUGBE database, 47% of participating sites routinely started a high-dose PPI before endoscopy.17 These uncertainties and variation in practice suggest that further research is needed to evaluate the role of these agents prior to endoscopy.

More convincing evidence of benefit comes from a meta-analysis of RCTs on the use of the anti-fibrinolytic agent tranexamic acid (TXA) for patients with UGIB. TXA reduces mortality in patients with traumatic bleeding and is used widely in elective surgical settings to reduce the need for blood transfusion.20 A meta-analysis of seven RCTs comparing TXA with placebo in 1754 patients with UGIB found a 39% relative reduction in mortality in patients prescribed TXA compared with controls (RR 0.61, 95% CI 0.42 to 0.89).21 However, there is limited evidence of its use in current practice. In the UK audit of acute UGIB, only 18 of 6750 patients were prescribed TXA,14 with no pertinent recorded data in published manuscripts of the Canadian RUGBE9 and Italian PNED studies.12 ,13 The disparity between the evidence base and its use in practice may reflect uncertainties about the generalisability of the published trials, as many were conducted prior to the era of PPI and routine endoscopic therapy. The increasing evidence base for this agent in other settings of haemorrhage warrants a re-evaluation of its efficacy in UGIB in future studies.

Finally, the quality of endoscopy can be adversely affected by poor visibility in patients requiring urgent endoscopy with UGIB due to obscuring blood in the gastric lumen. Pre-endoscopic administration of erythromycin or metoclopramide has been shown in a meta-analysis of five RCTs to reduce the need for repeat endoscopy in patients with UGIB (OR 0.55, 95% CI 0.32 to 0.94).22 This is an important adjunct to treatment with regard to the ability to make a diagnosis, apply definitive treatment and avoid unnecessary exposure of patients to repeat procedures; recent meta-analyses have even suggested a resultant decrease in hospital stay, so it should be used routinely in higher risk patients who are likely to have blood in their stomach, whether suspecting a non-variceal or variceal source.22 ,23 Intravenous erythromycin prior to endoscopy has also resulted in cost savings in cost-effectiveness analysis.24 Although data are not available in published papers in the aforementioned registries, the relative lack of this information may suggest that it was not in routine use at the time the pragmatic data were collected. Future studies should therefore specifically record its use, especially in higher risk patients.

Endoscopy within 24 h

All patients presenting with UGIB should ideally undergo endoscopy within 24 h of presentation, performed by endoscopists skilled in all modalities of haemostasis and ideally in appropriate clinical areas with skilled support staff available.8 Although no study has been able to show a direct reduction in mortality through performance of earlier endoscopy, the available data from observational studies favour a reduction in length of hospital stay, recurrent bleeding and need for surgery,25 ,26 with decreased mortality in subgroup analysis of patients at particularly high risk of a negative outcome.27 However, in the UK audit only 50% of all patients underwent endoscopy within 24 h of presentation (a figure which had not improved from the previous audit 14 years earlier), probably related to the lack of 24 h provision of emergency endoscopy in 48% of hospitals at the time of the audit in 2007.28 Although these figures were better in the Italian PNED-1 (78%), PNED-2 (81%) and the Canadian RUGBE registry (76%), longitudinal data from Canada showed that only 56% in the subsequent REASON study underwent endoscopy within 24 h.29 There is room for considerable improvement in this aspect of care delivery and this could be used as a key parameter to monitor standards and organisation of care delivery in future studies of UGIB.

Provision of 24/7 access to emergency endoscopy is an important prerequisite to performing endoscopy within 24 h for all patients. This is especially important to achieve in those countries which have demonstrated evidence of a ‘weekend effect’ for mortality following UGIB, including the USA30 ,31 and Wales.32 In contrast, no evidence of a weekend effect for mortality was found in Hong Kong33 or in the UK,34 although in the latter case this could have been related to goodwill as a similar degree of out-of-hours endoscopy was performed in those hospitals without formal rotas as in those with rotas. Nevertheless, the deficiencies in 24 h access to emergency endoscopy have led to the delivery of a key initiative in the UK. ‘Scope for improvement’ is a toolkit endorsed by multiprofessional bodies with recommendations to enable hospitals to support redesign of services to facilitate 24/7 management of patients developing UGIB in order to help bridge the gap between what is recommended in guidelines as optimum management for UGIB and what is actually delivered through existing models of care.35 Its issued recommendation as to the performance of endoscopy earlier than within the initial 24 h among high-risk patients, however, remains an area for further studies in the absence of current supportive data.36

Endoscopic therapy for high-risk lesions and haemostasis training

All patients with HRS at endoscopy should receive endoscopic therapy. Although monotherapy with epinephrine is effective compared with medical treatment alone, there is overwhelming evidence of the benefit of adding a second modality of treatment in patients with bleeding ulcers and HRS of haemorrhage in terms of reducing the subsequent risk of rebleeding, surgery and mortality, independent of which second procedure is added to the epinephrine injection (injection of another agent, heater probe, electrocoagulation or clips).37 ,38 Despite the strong evidence base, there are surprising shortfalls in translating this into clinical practice. In the RUGBE database, 72% of patients with HRS received any modality of endoscopic therapy with only 34% of those with HRS receiving dual endotherapy.9 Similarly, in the UK audit, 74% of those with HRS received any endoscopic therapy (a figure which was similar to the previous audit in 1993), with just 38% receiving dual endoscopic therapy.28 Although in both the PNED-1 and PNED-2 studies in Italy all patients with HRS received endotherapy, only 8.3% were reported to have received dual therapy in PNED-1,12 rising to 35% in PNED-2.13 Practice appears to be improving in Canada with such rates having risen to between 72% and 82%.11

Despite the strong evidence base from RCTs together with dissemination in international guidelines,8 it is unclear why such deficiencies exist in the application of endoscopic therapy (and specifically dual endoscopic therapy) in patients with high-risk lesions, and the rationale can be difficult to ascertain from retrospective studies. Although this could be explained in part by technical challenges in applying therapy (eg, due to the position of an ulcer), adequate training in endoscopic haemostatic techniques may play a role. Endoscopic haemostasis is often a challenging and unnerving procedure in the setting of an unstable patient, therefore adequate supervision and training can be difficult to provide to trainees compared with elective diagnostic procedures. The basic skills required by all endoscopists treating patients with NVUGIB include injection therapy, coagulation and haemoclip application, and there is evidence that simulator training for trainees can provide a useful adjunct to on-the-job training.39 The use of endoscopic therapy, as well as specifically dual endotherapy in lesions with HRS, seems a key parameter upon which longitudinal studies should monitor improvements in care.

Proton pump inhibitors after endoscopic therapy

There is strong evidence from RCTs supporting the efficacy of high-dose proton pump inhibition in patients with HRS who have received endoscopic haemostasis. Most of these trials have used either continuous intravenous infusions or high-dose boluses for 72 h on the basis that most rebleeding is known to occur within the first 3 days. A meta-analysis including 5792 patients across 24 RCTs showed that PPIs significantly reduced the incidence of rebleeding and the need for surgery compared with placebo.40 In practice there is poor adherence to the guidance on the use of high-dose intravenous PPI as an adjuvant to endoscopic haemostasis, as is currently recommended based on the strongest evidence, although increasing poorer quality data suggest that lower doses and other routes of administration may also be efficacious.41 In the Canadian REASON-2 study, less than 10% of patients who underwent endotherapy for HRS also received the full complement of a 72 h continuous PPI infusion.11 The premature discontinuation may reflect the aforementioned lack of certainty among clinicians as to the necessity of an intravenous PPI bolus followed by a full 72 h infusion. Further studies are therefore needed to evaluate shorter regimes, lower doses and oral formulations in this setting.

On which parameters should we monitor future improvements in care?

A detailed list of suggested outcomes as they apply to the performance of RCTs can be found elsewhere.42 A separate list of recorded parameters needs to be drawn up for observational studies and we have recommended some variables to collect. Perhaps the most surprising shortcoming in existing care highlighted above is the relative lack of use of endoscopic therapy in all patients found to have lesions with HRS at endoscopy and the even more limited use of dual endoscopic therapy in these patients despite the strong evidence base of these modalities leading to reductions in rebleeding. Given the persistently high rates of rebleeding, this aspect of care seems a key parameter upon which future studies should monitor the quality of endotherapy delivered to patients with UGIB. Similarly, the proportion of patients in whom endoscopy is achieved within 24 h would seem a sensible performance indicator to monitor improvements in processes of care. Demonstration of adequate skills in endotherapy is important for all endoscopists performing emergency endoscopy for UGIB, and skills assessment could form part of accreditation in training programmes.

Controversies and areas for future research

Rebleeding rates remain stubbornly high, in the order of 10–15%. Ongoing and future areas of research should therefore focus on improving this endpoint, which is a strong predictor of mortality. With regard to pharmacological treatment, the optimum dosing, route and timeframe for administration of PPIs after endoscopic haemostasis of high-risk lesions needs further evaluation. The usefulness of antifibrinolytics as clot stabilising agents and adjuncts to current therapy needs further evaluation. With regard to endotherapy, emerging technologies may play an important role and may not necessitate the same degree of technical ability as is required in some existing modalities.43 Finally, a better understanding is needed of the increasing role of radiological embolisation in the current management algorithm for patients who experience further bleeding.