Over the past decades, stress-related bleeding has become extremely uncommon in intensive care unit (ICU) patients [1]. Apart from pharmacologic approaches for stress ulcer prophylaxis (SUP), advances in the care of critically ill patients, such as optimal fluid resuscitation to maintain hemodynamic stability and thus improve splanchnic perfusion, and early provision of enteral nutrition (EN), may contribute to this observation [2–4]. Although recommended only in patients on mechanical ventilation or coagulopathy, patients with traumatic brain injury or major burns, or those with ≥ 2 risk factors [5, 6], SUP is still being used in nearly 90% of ICU patients, despite lack of an accepted indication in the majority [7–9]. Furthermore, SUP is often continued in these patients until clinical improvement, or even after transfer to the general ward [1, 10]. However, SUP is not without risks. The extensive use of SUP has been demonstrated to be associated with a higher rate of hospital-acquired pneumonia (HAP) due to loss of the protective bacteriostatic effect of gastric acid [4, 11]. Meanwhile, concomitant treatment of SUP and broad-spectrum antibiotics has also contributed to higher risks of Clostridium difficile infection [12, 13]. Thus, selection of potentially high-risk patients who may benefit from SUP while avoiding unnecessary use in others is important.

Some earlier studies reported that EN alone might provide sufficient prophylaxis against stress-related gastrointestinal (GI) bleeding [3, 14]. In animal models, enteral feeding is documented to increase GI blood flow and provide protection against GI bleeding [15, 16]. In a prospective, open-label trial, continuous EN was shown more likely than proton pump inhibitors (PPIs) or histamine 2 receptor antagonists (H₂RAs) to raise gastric pH to above 3.5, suggesting that EN might be more effective in preventing GI bleeding than pharmacologic SUP [17]. Although several recent systematic reviews have comparatively evaluated pharmacologic agents for SUP, few of these studies have specialized in patients received EN [4, 18–20]. In 2010, one meta-analysis comparing H₂RAs to placebo or no prophylaxis for SUP looked into a subgroup of enterally fed patients. In this subgroup, SUP did not decrease the risk of bleeding, and in contrast led to more episodes of hospital-acquired pneumonia (HAP) and higher mortality rate [4]. However, these findings were based on an evaluation of only 262 patients in three randomized controlled trials (RCTs) (three trials in GI bleeding, two trials in HAP and mortality), which were published between the years 1985 and 1994 and compared H₂RAs with placebo [21–23]. In addition, two out of the three RCTs were unblinded [21, 22], and some of potentially important outcomes to clinicians or patients, including duration of mechanical ventilation, incidence of C. difficile infection, ventilator-associated pneumonia (VAP) and length of ICU stay were not considered in this meta-analysis.

Therefore, in order to address these limitations, we sought to expand the previous meta-analysis by adding relevant RCTs published between 1994 and 2017, and including any prophylaxis regimens. We reviewed these RCTs to determine if there are differences between pharmacologic SUP and placebo or no prophylaxis in enterally fed patients in terms of stress ulcer-related GI bleeding, and other clinical outcomes.

Our meta-analysis showed that, among ICU patients receiving enteral feeding, pharmacologic SUP exerted no impact on the risk of GI bleeding, overall mortality, C. difficile infection, duration of MV and length of ICU stay, but led to an increased risk of HAP.

In this updated meta-analysis, we found that there was no added benefit with concomitant pharmacologic SUP in GI bleeding once patients were receiving enteral feeding. This finding expanded on the earlier meta-analyses to provide better evidence for pharmacologic SUP in enterally fed patients in the ICU [4]. First, our meta-analysis had a larger sample size than the previous meta-analyses as it included four RCTs published between 2016 and 2017, with more power to assess this effect. Second, the subgroup and sensitivity analyses based on various clinical characteristics did not significantly alter our main findings. Finally, we further evaluated other related important outcomes (e.g., duration of mechanical ventilation, overall mortality, and length of stay in the ICU) and found no difference between groups, thus providing evidence of the robustness of our results.

Apart from calorie delivery, EN had also been found to protect against stress-related GI bleeding [4, 14–16]. EN might mitigate macroscopic ulceration by optimizing mucosal energy, intramucosal pH [17, 30] and regional distribution of gastrointestinal blood flow [31, 32]. In addition, EN was able to reduce GI inflammation secondary to reperfusion injury. To date, there had been no RCTs comparing GI bleeding rates in critically ill patients receiving EN versus no EN. Several case series involving a total of 749 patients reported that enterally fed critically ill patients in the absence of pharmacologic SUP was associated with bleeding rates of 8.4–8.8% [3, 33]. In our study, we also found a similar bleeding rate (7.5%) in patients receiving EN alone. In comparison, Marik et al. reported a GI bleeding rate of 15.8% in the subgroup of patients without EN or any pharmacologic SUP [4]. A similar incidence (17.5%) was also identified in a recent meta-analysis in unfed patients [19]. These data suggest a potential role of EN against stress-related GI bleeding, therefore questioning any added benefits of pharmacologic SUP in patients already receiving EN.

Our results indicated that pharmacologic SUP in enterally fed patients is associated with higher incidence of HAP, which is consistent with the previous meta-analysis [4]. The reason might be that concomitant EN and pharmacologic SUP would result in a significantly higher pH than either intervention alone [17]. However, we should interpret this finding cautiously. First, the definitions of HAP varied across included studies, with the incidence ranging from 8.4 to 52.9% [18, 21–23, 27–29]. Second, the included studies had spanned a period of more than three decades, when co-interventions had been developed and quality improvement approaches such as guidelines for HAP/VAP prevention had been introduced and updated [34]. This, to some extent, might affect the accurate evaluation of the effects of SUP. In fact, the significant increase in pneumonia was mainly caused by the two earlier trials [21, 22], but not the newly published RCTs [18, 27, 28]. When only the three RCTs that focused on VAP were considered, no differences were found between groups [18, 27, 28]. Finally, we did not find significant differences in terms of other secondary outcomes (e.g., mortality, duration of mechanical ventilation, or length of stay in the ICU). Recently, more and more attention had been paid to the possible association between the SUP strategy and enteric infections, particularly C. difficile [13]. Several epidemiological investigations and meta-analyses had demonstrated an increased risk of C. difficile infection in patients under a SUP strategy. In addition, studies suggested that PPIs were more strongly associated with this enteric infection than H₂RAs [12, 13]. However, these results could not place sufficient weight on RCT evidence. It was noteworthy that no study had investigated the effect of EN on C. difficile infection. Our results suggested that in enterally fed patients the rate of C. difficile infection was similar in the SUP and non-SUP groups (1.9% vs. 2.0%). The relatively small number of events may account for these negative results. Therefore, further well-designed, large RCTs are warranted to focus on this topic, as the detrimental outcomes of these serious infections may outweigh the benefit of SUP.

Recently, there has been growing interest in PPIs as a means of SUP. For example, all four additional RCTs included in our meta-analysis compared the effect of PPIs with placebo. Moreover, PPIs were also increasingly prescribed as the primary SUP agent, ranging from 39.6 to 70% in critically ill patients [8, 35]. In a recent international survey, PPIs were the most comment agent (66%) used for SUP [1]. This might be due to the superiority of PPIs in reducing GI bleeding, as suggested by several meta-analyses [20, 36], and the recommendations of the Surviving Sepsis Campaign [37]. Despite the widespread use of PPIs, the effect of concomitant EN in this SUP procedure is rarely evaluated. In our study, only four included RCTs focused on this topic, and pooled results suggested no benefit or harm associated with PPIs. Though limited by the small sample, our data may, at the very least, encourage clinicians to reevaluate their practice in prescribing prophylactic PPIs in critically ill patients. As a matter of fact, several ongoing RCTs comparing PPIs with placebo in ICU patients with high risk of GI bleeding may provide more convincing evidence in the future [38–40].

Our study has some limitations. First, only seven studies were included in our analysis, and most of them had a sample size of less than 200 [18, 21, 23, 28, 29], which would more likely result in overestimation of effect size. Thus, further studies in large cohorts are needed to validate our findings. Second, there were differences among included trials with regards to the adopted definition of GI bleeding, timing and duration of EN, and patient intolerance of EN, which might lead to the observed heterogeneity, and therefore compromise the robustness of our findings. Third, the uneven distribution of different underlying diseases among included studies might also exert a prognostic value. We planned to perform subgroup analyses to explore studies based on such diversities, which was hampered by nsufficient data. Fourth, although predefined subgroup analyses had been performed, some results of subgroups should be interpreted with caution due to small number of patients. Finally, we had not pre-published this updated meta-analysis protocol in a registry.

In summary, based on available data, our results demonstrate that in ICU patients receiving EN, pharmacologic SUP offered no beneficial effect on the incidence of GI bleeding and other clinically important outcomes. Large-scale, well-designed RCTs will be needed to confirm our findings.