Absence of Helicobacter pylori is not protective against peptic ulcer bleeding in elderly on offending agents: lessons from an exceptionally low prevalence population

Introduction

Non-variceal upper gastrointestinal (GI) bleeding remains a prevalent condition and its mortality hardly change despite a declining trend of peptic ulcer disease and improvement in therapeutic approaches. The reported incidence from North America and Europe was 20–60/100,000 populations but data from Asia were unfortunately scarce and variable (Sung et al., 2011). A recent study from Thailand reported an incidence of 152.9/100,000 population (Sangchan et al., 2012) and data from East Malaysia (State of Sabah), available only in abstract, reported an incidence of 72/100,000 population (Cheng et al., 2001). Reports from two tertiary hospitals in central Peninsular Malaysia indicated an overall low prevalence of non-variceal upper GI bleeding among the ethnic Malays (Lim et al., 2003; Lakhwani et al., 2000; Lee, Mahendra Raj & Graham, 2013).

There is a reducing trend of peptic ulcer disease observed within Asia, and this is largely a result of reducing prevalence of H. pylori infection. This trend is likely to continue into the future and a time will come when H. pylori joins the ranks of smallpox and polio (Graham, Yamaoka & Malaty, 2007). The population in the north-eastern region of Peninsular Malaysia (state of Kelantan), that consists of 90% ethnic Malays, had a seroprevalence of Helicobacter pylori (H. pylori) infection of only 4.2% among 496 blood donors and 4.8% among 921 patients attending a health screening clinic (Uyub et al., 1994). The H. pylori infection rate reported from gastric biopsies was 20% in duodenal ulcer, 21.2% in gastric ulcer, 16.7% in duodenal erosion and 17.1% in gastric erosion (Raj et al., 2001). The incidence of peptic ulcer perforations within the region from 1991 to 1992 was only 1.5/100,000 person-years (Uyub et al., 1994).

The exceptional low prevalence of H. pylori in the population from north-eastern region of Peninsular Malaysia provides us with an opportunity to contemplate the future without the infection. Our study aimed to determine the risk and clinical outcomes of acute non-variceal upper GI bleeding in this population with a low prevalence of H. pylori infection. The association between clinical characteristics, risk factors and treatment given with risk and bleeding outcomes was also assessed.

Methods

Results

Primary outcome

The majority of patients were of low risk on admission with 85.2% (213/250) of patients had a Rockall score ≤ 5 and a mean Rockall score of 4.4. There were 3.6% (9/250) in-hospital mortality, 9.6% (24/250) recurrent bleed and 4.0% (10/250) of patients who subsequently required surgery. A higher Rockall score in this population was associated with increased in-hospital mortality (mean score 7.0, P < 0.001), recurrent bleeding (mean score 5.1, P = 0.01) and need for surgery (mean score 4.8, P = 0.01). A Rockall score > 5 was significant in predicting recurrent bleeding in this population but only with area under curve or AUC of 0.6 (95% CI: 0.5–0.7, P = 0.04) (Fig. 1).

Discussion

In this population starting with an exceptionally low prevalence of H. pylori infection, the incidence of 2.2/100,000 person-years of non-variceal upper GI bleeding was also notably low. Peptic ulcer bleeding was the most common cause with an incidence of 1.8/100,000 person-years and this was almost similar to previously reported peptic ulcer perforation of 1.5/100,000 person-years (Uyub et al., 1994). The elderly in this population were more susceptible to non-variceal bleeding, especially from DUs and had a higher risk score and concomitant co-morbidities, in keeping with recent observation in H. pylori-eradicated populations (Domon et al., 2012).

The Rockall score > 5 had been shown to be useful in predicting recurrent bleeding in our population but the AUC result suggests that it may be less satisfactory. This might be related to the threshold of 5 that was initially chosen in the current study. Since many patients admitted to referral institutions were likely to have high risk endoscopic features (stigmata of recent bleed) therefore a threshold of 5 seems reasonable in order to discriminate high risk from the low risk patients, as previously reported (Bessa et al., 2006; Camellini et al., 2004). However, our results reported a lower rate of high risk features, which would then have resulted in a less accurate AUC. Hence, a threshold below 5 may have been preferable in our population and this merits further analysis. A number of recent studies have similarly questioned the accuracy of Rockall scores in different populations and settings (Wang et al., 2013; Custódio Lima et al., 2013). Further prospective studies are needed to verify their findings.

Our study shared similar baseline characteristics with Vreeburg et al. (1999) including definition for mortality but our results suggest a better prediction of recurrent bleeding rather than in-hospital death. A higher rate of recurrent bleeding observed in our population as compared to in-hospital mortality might explain this discrepancy. The low hemoglobin and urea levels indicate a minor bleeding risk in general, compatible with the overall low risk score observed in this population. The generally low risk score in this population would not, however, allow one to decide for the need of therapeutic endoscopy. The Blatchford score may have been more useful in this regard (Pang et al., 2010).

Among the variables described in Rockall score, SRH stood out as the most predictive of high risk and in-hospital mortality. The presence of SRH is of greater significance in H. pylori-associated bleeding GUs than DUs (Chang-Chien et al., 1988). In our study population, more than half of upper GI bleeding was a result of GUs with only 20% due to DUs. However, GUs were more likely Forrest III lesions (57.2%) but had relatively benign outcomes. In contrast, DUs, while less common, and were more likely Forrest I and II lesions, but there was significant associations between DUs with all studied outcomes. Previous studies have also similarly observed that H. pylori-negative DUs were more likely to bleed and were more common among the elderly population with risk factors (Gisbert & Calvet, 2009; Chu et al., 2005). These studies were limited by false negative results for H. pylori, but our study population did not suffer from this limitation (Gisbert & Calvet, 2009).

Septicaemia, while not a variable in the Rockall score, was also highly predictive of high risk and in-hospital mortality, similarly reported by Zimmerman and others (Zimmerman et al., 1995; Afessa, 1999). In the original Rockall validation study, pneumonia which was associated with septicaemia was included in the model but not in the complete model (Rockall et al., 1996). Our study suggests that septicaemia, if present, should be considered as a major co-morbidity and should be given a score of 2. An elevated creatinine was predictive of increased risk of recurrent bleeding in the multivariable analysis, in agreement with Zuckerman and others (Zuckermann et al., 1985; Sood et al., 2012). Patients with ischemic heart disease, the most common co-morbidity in our study, were also frequently impaired in their renal function (Shalev et al., 2012).

Non-variceal upper GI bleeding was associated with more adverse outcomes in the current study with a mostly elderly population, and an almost absence of H. pylori infection, and in the presence of offending agents including aspirin, NSAIDs and warfarin. In a study from Japan, the usage of aspirin and NSAIDs was not associated with a serious outcome in GU bleeding (Ishikawa et al., 2008) but the role of H. pylori infection was not addressed. Recent studies found that patients with H. pylori-negative peptic ulcers and who took aspirin were more likely to have a higher bleeding risk (Kang et al., 2011; Chan et al., 2013; Hernandez-Diaz & Garcia Rodriquez, 2006). NSAIDs and aspirin in combination, rather in isolation, would have produced the greatest risk, and this might explain why our multivariate analysis did not have any association (Hreinsson et al., 2013; Lee & Noridah, 2013). Further studies are needed to determine the significance of our findings since these results may have a long term impact in the H. pylori-eradicated populations.

Endoscopic interventions were carried out in only a third of patients with high risk scores and this implied that the Rockall score is not useful in selecting patients who require interventions. Endoscopic therapy with adrenaline only was associated with a four-fold risk for recurrent bleeding and a ten-fold risk for surgical intervention based on the multivariable analysis; figures that were similar to Levin et al. (2012). This indicates that adrenaline alone is unlikely to be sufficient when endoscopic interventions are needed and is not in line with our current practice or guidelines (Chung et al., 1997; Chung et al., 1999).

The need for surgery was not an outcome initially included during the validation study of the Rockall score, however surgical intervention is frequently sought in the setting of failed endoscopic therapies. In the current study, the need for surgical intervention of 4.0% was relatively similar to the rate of in-hospital mortality of 3.6%. Previous study indicated an overall mortality of 34.1% in patients with upper GI bleeding requiring surgery (Czymek et al., 2012). Epigastric pain, predictive of the need of surgery, may be a sign of impending perforation, and should be carefully assessed especially in this largely rural-majority population who often present late in their course of disease.

Study limitations should be mentioned. There was only one other centre in the region managing upper GI bleeding, and this might affect our incidence calculation slightly but our institution is a primary referral institution with a reliable GI database. The rate of H. pylori infection was underestimated in this study for several reasons. One was the use of rapid urease test, which had been shown to be unreliable during the acute phase of ulcer bleeding (Schilling et al., 2003; Lee et al., 2000). Due to this reason, based on the discretion of the managing endoscopist, only 35.6% of patients were tested. Of those tested negative with the urease test, histology was reliable in confirming the absence of H. pylori without other additional tests. On the other hand, this limitation on H. pylori would not have affected our current study since we have already published extensively on the exceptional low prevalence of H. pylori in the region (Lee, Mahendra Raj & Graham, 2013; Uyub et al., 1994; Raj et al., 2001).

So what would be the future in patients having non-variceal upper GI bleeding in the absence of H. pylori? We can conclude that based on our population with an exceptional low prevalence of H. pylori infection and also peptic ulcer disease, acute non-variceal upper GI bleeding is also of low incidence, similar to peptic ulcer perforation rates. An absence of H. pylori infection may not however reduce the risk of peptic ulcer bleeding in the presence of risk factors especially offending drugs in an elderly population.

Supplemental Information