Cook and associates [13] examined the risk factors for gastrointestinal bleeding in patients entering the ICU. The incidence of clinically important gastrointestinal bleeding was only 0.1% for patients without respiratory failure or coagulopathy, whereas patients with one or both of these risk factors had a 3.7% incidence of substantial hemorrhage. These observations suggest that most patients entering the ICU will not benefit from prophylaxis for stress-related gastritis. Whether prophylaxis can reduce the incidence of bleeding for patients at high risk was not investigated in that study.

In addition to the issue of efficacy, concerns exist regarding the safety of prophylaxis. Specifically, investigators suggest that some regimens may be associated with an increased incidence of nosocomial pneumonia [7, 14, 15]. If the risk for stress-related hemorrhage is so low that prophylaxis benefits few patients and if prophylaxis increases the risk for concomitant illness, then the role of prophylaxis for stress-related bleeding should be reevaluated.

We completed a randomized, controlled, blinded study to determine the role of prophylaxis for stress-related gastritis for patients entering a medical ICU. Our specific goals were to determine the incidence and outcome of clinically important stress-related hemorrhage for patients in a medical ICU and the safety and efficacy of cimetidine and sucralfate prophylaxis.

Methods

Top Methods Results Discussion Author & Article Info References

Study Patients

Henry Ford Hospital, a 900-bed tertiary-care teaching institution, has 28 adult medical ICU beds, excluding those in the coronary care unit. All patients 18 years or older who were admitted to the medical ICU between 1 February and 25 November 1992 were screened for study eligibility. Exclusion criteria were 1) expected stay of 24 hours or less; 2) evidence of gastrointestinal bleeding [hematemesis, vomiting of "coffee grounds," hematochezia, or melena] at the time of admission to the ICU; 3) treatment with antacids, H₂-receptor antagonists, sucralfate, or omeprazole during the 24 hours before entering the ICU; 4) use of nonsteroidal anti-inflammatory agents, systemic anticoagulants, or thrombolytic agents during the 7 previous days; 5) surgery requiring general anesthesia during the previous 2 weeks; 6) closed head injury or clinical evidence for increased intracranial pressure; 7) grade 4 hepatic encephalopathy; 8) esophageal or gastric surgery in the previous year; 9) history of gastrointestinal bleeding during the previous year; and 10) pregnancy or lactatation. Patients with several ICU admissions during the study period were randomly assigned to a therapy protocol only once. This protocol was reviewed and approved by the Henry Ford Hospital Institutional Review Board. Informed consent was obtained from the patient or from legally authorized representatives when the patient could not provide consent.

Therapy

Random assignment to one of the two study treatments or control group and commencement of therapy occurred within 6 hours of ICU admission. Randomization was by sealed envelope using the permuted block design [16]. Control patients did not receive antacids, sucralfate, omeprazole, or H (₂-receptor) antagonists. Patients assigned to receive sucralfate were given 1 g of the medication orally or as a suspension through the nasogastric tube every 6 hours. The nasogastric tube was clamped for 1 hour after sucralfate administration. Patients receiving cimetidine were given a 300-mg intravenous loading dose followed by continuous intravenous infusion according to the creatinine clearance: more than 50 mL/min, 900 mg/d; 20 to 50 mL/min, 600 mg/d; and less than 20 mL/min, 300 mg/d. The cimetidine dose was titrated to maintain gastric pH 4.0. If two consecutive gastric pH values were less than 4.0, the dose was increased by the following amounts based on creatinine clearance: 300 mg/d, 200 mg/d, and 100 mg/d. The maximum allowable cimetidine doses for the patients grouped by renal function were 2400 mg/d, 1600 mg/d, and 800 mg/d. Nasogastric tubes were not required as part of the protocol, but gastric pH was checked every 2 hours for patients with these tubes (using S/P pH Indicator Strips, Baxter Health Care, McGaw Park, Illinois).

Substantial gastrointestinal hemorrhage required the presence of any of the following: 1) persistent hematemesis (red blood or guaiac-positive "coffee grounds") that did not clear with 1.5 L saline lavage; 2) 3-point decrease in hematocrit during 24 hours accompanied by red blood or guaiac-positive "coffee grounds" material that cleared with lavage, or melena, or three guaiac-positive stools without evidence of lower gastrointestinal bleed; and 3) any unexplained 6-point decrease in hematocrit during a 48-hour period. This last criterion was added as a safety measure because some patients would not receive prophylaxis.

Data Collection

On admission to the ICU, the following data were collected: age; sex; admission diagnoses; APACHE II score [17]; vital signs; result of chest roentgenogram; arterial blood gases; and hematologic, biochemical, and coagulation laboratory values. Data collected daily during the ICU stay included vital signs, fluid intake and output, presence of blood in the stool or nasogastric aspirate, hematocrit value, serum creatinine concentration, coagulation values (prothrombin time, partial thromboplastin time, and platelet count), results of chest roentgenogram, gastric pH values, signs and symptoms of gastrointestinal hemorrhage, number of units of packed red blood cells transfused, medications, volume of total parenteral or enteral nutrition, and possible adverse drug reactions [18-20]. Possible central nervous system reactions to cimetidine were defined as anxiety, confusion, delirium, depression, hallucination, headache, or somnolence [21]. The presence of putative risk factors for stress-related hemorrhage [22-24] was recorded at entry into the study and daily during the ICU stay (Table 1).

Outcomes

The primary study end point was substantial hemorrhage from stress gastritis. Information regarding hematocrit, Hemoccult status of stool and nasogastric aspirate, and volume status was presented daily to two investigators (RF and RSB) who were blinded to therapy. These investigators determined whether patients satisfied the criteria for substantial hemorrhage. All patients who met the criteria received endoscopy within the next 24 hours performed by one of the two investigators. Treatment was considered to have failed only in patients who were diagnosed with stress-related hemorrhage by the endoscopy. After diagnosis of stress-related hemorrhage, the patients' study medication was stopped and they were treated with continuous intravenous infusion of cimetidine and with blood transfusions, if necessary. This treatment protocol is the standard in our medical ICU.

Adverse drug reactions were recorded, and the Adverse Drug Reaction Probability score was calculated [26]. In the appropriate clinical situation, patients were rechallenged with the same medication within 24 hours. If the study medication was discontinued, prophylaxis was continued at the discretion of the primary physician.

Nosocomial Pneumonia

Diagnosis of nosocomial pneumonia in the medical ICU required all of the following: 1) chest roentgenogram obtained 72 hours or more after ICU admission that showed a new and persistent infiltrate; 2) fever, leukocytosis, or both; 3) purulent tracheobronchial secretions; 4) Gram-stained sputum showing more than 25 polymorphonuclear leukocytes and fewer than 10 squamous epithelial cells per low-power field; and 5) recovery of an accepted nosocomial pathogen from sputum culture.

Additional end points were total transfusion requirement, recurring hemorrhage, duration of hospitalization, death in the ICU, and outcome of hospitalization (death or discharge).

Statistical Analysis

We estimated sample size to provide 80% power to detect a 75% reduction in bleeding rate, that is, a 12% bleeding rate for the control group compared with a 3% rate in either of the two treatment groups. We used an value of 0.05 (two-tailed) adjusted for the comparison of the control group with each of the prophylaxis groups. As a result of these assumptions, 160 patients were needed in each of the three groups.

Results are expressed as mean ±SD, or median and interquartile range for lengths of stay. We compared baseline differences using chi-square tests and analysis of variance [27]. The primary data analysis compared the incidence of bleeding and death between the control group and each of the prophylaxis groups. We confirmed the results of these tests with more detailed analysis using Cox regression [28]. Differences in incidence are expressed as relative risks, with their corresponding 95% CIs. To compare lengths of stay, we calculated median differences and 95% CIs [29].

Safety and Quality Assurance

To guarantee patient safety, an independent monitoring committee of two physicians and a biostatistician was established. Meetings of the committee were planned for points approximately one third and two thirds through the study period. We authorized this committee to recommend study termination based on patient safety or statistical concerns. The committee used the method of Lan and DeMets [30] to monitor the primary study outcomes. They performed stochastic curtailment analyses using the "B-value" technique [31] for the outcomes of bleeding or death. They computed the probability of a significant and of a nonsignificant final result assuming that the rest of the data showed the difference expected under 1) the alternative hypothesis, 2) the null hypothesis, or 3) the currently observed incidence.

Results for 291 patients were available for the second committee meeting on 24 November 1992. The incidences of endoscopically confirmed stress-related bleeding were 6 of 95, 4 of 98, and 5 of 98 for the control, sucralfate, and cimetidine groups, respectively. Assuming that the alternative hypothesis held for the remaining observations, the conditional power of a significant result was 6.4% for sucralfate compared with control and 0.3% for cimetidine compared with control. Termination of the study was recommended because of low conditional power [32]. Thirteen patients had been randomized and were still in the ICU when the study was terminated. They were followed during their ICU stay according to protocol.

Quality control for data collection involved comparing randomly chosen data values with the corresponding patient charts. Data entry was by double entry with appropriate editing performed.

Results

Top Methods Results Discussion Author & Article Info References

During the study, 877 patients were admitted to the ICU 922 times. Three hundred four admissions were randomized, with four subsequently excluded from analysis. (One patient died 2 hours after admission. Three patients were randomized on two different occasions, and the second admissions were excluded.) The common exclusion criteria and the percentage excluded for that reason were gastrointestinal hemorrhage at the time of ICU admission or in the previous year (48%); use of antacids, sucralfate, H₂-receptor antagonists or omeprazole during the 24 hours before ICU admission (37%); expected stay less than 24 hours (15%); use of nonsteroidal, anticoagulant, or thrombolytic agents (15%); and inability to obtain informed consent (11%). Most excluded patients met more than one of these criteria.

One hundred patients were randomly assigned to each of the treatments. The groups were similar with regard to age, sex, percentage of participants admitted from the emergency room, severity of illness, and admission diagnoses (Tables 2 and 3). Patients often had more than one reason for ICU admission. The mean APACHE II scores for the control, sucralfate, and cimetidine groups were 16.5 ±6.9, 16.8 ±6.9, and 18.0 ±8.0, respectively. Approximately one third of the patients in each group had APACHE II scores greater than 20.0. Mechanical ventilation for more than 24 hours was needed by 65% to 76% of patients in the different groups. These demographic and clinical characteristics are similar to those of the 143 patients admitted to the ICU but not entered into the study because they were taking one of the study medications or because informed consent could not be obtained (data not shown).

Table 4 shows the distribution of putative risk factors for stress-related gastritis. Risk factors were recorded until the onset of hemorrhage for patients who bled or at any time during the entire ICU stay for those who did not bleed. The distributions of risk factors at ICU admission were similar, except for a higher incidence of the sepsis syndrome in the cimetidine group (P = 0.04). After inclusion of risk factors acquired in the ICU, the distribution of this and the other risk factors was similar among the three groups. The mean number of putative risk factors per patient was 2.0 ±1.5 for the controls, 2.1 ±1.7 for the sucralfate group, and 2.5 ±1.8 for the cimetidine group (P = 0.22).

Figure 1 shows the frequency distribution of risk factors per patient. Thirty-two percent of the study population had two risk factors. The proportions of patients with 3 or more risk factors were 32%, 31%, and 34% in the control, sucralfate, and cimetidine groups, respectively. There were no statistically significant differences in the frequency distributions among the three groups (P > 0.05).

View larger version (20K): [in this window] [in a new window]   Figure 1. Number of risk factors per patient in the three study groups. Table 1 lists risk factors. There were no statistically significant differences in the distributions among the three groups (P > 0.05).

The proportions of gastric pH values 4.0 were 64%, 72%, and 88% in the control, sucralfate, and cimetidine groups, respectively. The gastric pH did not differ between the control and sucralfate groups, whereas there was a statistically significant difference between the cimetidine group value and that of the other two groups (P < 0.01).

The dose of cimetidine was reduced in patients with diminished renal function. For subjects with creatinine clearance rates less than 20 mL/min (n = 24), the mean cimetidine dose required to maintain gastric pH 4.0 was 390 ±223 mg/d. For patients with rates between 20 and 50 mL/min (n = 30) and those with rates greater than 50 mL/min (n = 46), the cimetidine doses necessary to maintain pH 4.0 were 735 ±367 and 1060 ±408 mg/d, respectively. The percentage of gastric pH values 4.0 was not significantly different (P > 0.05) among the cimetidine subgroups defined by renal function. Only one patient did not achieve adequate control of gastric pH despite reaching the maximal cimetidine dose, and this patient did not hemorrhage.

Most patients received enteral nutrition through a 10-Fr feeding tube. Enteral feeding was administered to 72% of control patients, to 69% of those receiving sucralfate, and to 57% of those receiving cimetidine (P = 0.06). In the control and sucralfate groups, 1% of patients received parenteral nutrition and 7% of the cimetidine group received this therapy (P < 0.015 compared with either of the other groups). The incidence of stress-related hemorrhage among patients receiving enteral or parenteral nutrition was not significantly different compared with the incidence in those who did not receive nutritional supplements (P > 0.05 for all comparisons).

Gastrointestinal Hemorrhage

Forty-three patients satisfied our criteria for significant hemorrhage: 11 patients met criterion 1, 13 met criterion 2, and 19 met criterion 3. The distribution of the criteria for hemorrhage was similar among the three treatment groups. Esophagogastroduodenoscopy was done in 34 of the 43 participants. The mucosa was examined in one patient at autopsy done 18 hours after death. Table 5 lists the findings. Sixteen patients had stress-related gastritis (6 controls, 5 receiving sulcralfate, and 5 receiving cimetidine). The relative risk for either prophylaxis group compared with control was 0.83 (CI, 0.26 to 2.64; P = 0.75). The median time from ICU admission to onset of stress-related hemorrhage was 5 days. There were no significant differences among the three groups, either by log-rank test or by Cox regression adjusting for respiratory failure and high-dose corticosteroid use.

Esophagogastroduodenoscopy showed erosions limited to the proximal stomach, except for a patient with erosive gastritis of body and antrum and another with involvement of the antrum only. Endoscopy showed evidence of active bleeding in five of the patients.

The transfusion requirements during the 48 hours after onset of stress-related hemorrhage were 1.2 ±1.4, 2.0 ±2.0, and 1.6 ±1.3 units of packed red blood cells for the control, sucralfate, and cimetidine groups, respectively (P = 0.41). One patient receiving sucralfate required a 4-unit transfusion. Five patients (1 control, 2 sucralfate, and 2 cimetidine) required no transfusion. No patient required therapeutic endoscopy, embolization, or surgery to stop the bleeding.

Five of the 16 patients with stress-related hemorrhage died. In none of them was bleeding the cause of death. A patient receiving sucralfate had cardiogenic shock and died 6 hours after the onset of hemorrhage. A patient receiving cimetidine and a patient in the control group had multiple-organ dysfunction syndrome unresponsive to therapy and died within 24 hours of hemorrhage. Two patients (1 in the sucralfate and 1 in the cimetidine group) died 4 and 13 days after bleeding as a result of underlying disease. These last 2 patients did not have recurrent hemorrhage.

The other 11 patients with stress-related bleeding survived and were transferred to the medical wards. In these cases, duration of bleeding was less than 24 hours, and bleeding did not recur during hospitalization.

Patients with stress-related hemorrhage had more risk factors than did patients who did not bleed (3.1 ±1.7 compared with 2.2 ±1.6, respectively; P = 0.025). Of these factors, only respiratory failure (P = 0.03) and high-dose corticosteroid use (P = 0.01) were associated with a statistically significant risk for stress-related hemorrhage.

In the subgroup consisting of patients with either respiratory failure or high-dose corticosteroid use (that is, the high-risk subsets), prophylaxis did not substantially reduce the incidence of stress-related hemorrhage. In the control group, 5 of 65 patients (8%) with respiratory failure had stress-related hemorrhage. For the sucralfate and cimetidine groups, the incidences were 5 of 72 (7%) and 5 of 76 (7%), respectively. Forty-one patients in the control group received corticosteroids, but only 3 hemorrhaged (7.3%). In the sucralfate group, 4 of 40 patients (10%) receiving corticosteroids hemorrhaged, and 4 of 37 (10.8%) in the cimetidine group bled.

Because coagulopathy has been identified as an independent risk factor for stress-related hemorrhage in some studies [13, 23], we analyzed our data to determine whether prophylaxis reduced the incidence of stress-related hemorrhage among patients with coagulopathy. Two of 20 patients in the control group with coagulopathy had stress-related hemorrhage. In the cimetidine and sucralfate groups, the incidences were 1 of 22 and 2 of 17, respectively (P > 0.05).

Nineteen patients who met our criteria for substantial hemorrhage did not have stress-related bleeding. Twelve patients (4 controls, 3 receiving sucralfate, and 5 receiving cimetidine) had a normal result of endoscopy, suggesting that the change in hematocrit was due to fluid shifts. None of these patients died or received transfusions. Seven patients bled from causes other than stress ulceration (Table 5). All 19 patients were discharged alive from the hospital.

Eight patients did not have endoscopy. Consent was not provided for five patients (one receiving sucralfate, three receiving cimetidine, one control) who satisfied criterion 1 or 2. Four of these patients had refractory multiple-organ failure and died of underlying diseases or because of withdrawal of life support within a few hours of hemorrhage. One patient with lymphoma and thrombocytopenia died of multiple-organ system failure 10 days after the bleeding episode. Three of the eight patients met criterion 3. One patient, receiving sucralfate, died 24 hours after withdrawal of life support. One patient in the control group had obstructing esophageal cancer that precluded passage of the endoscope. The third patient, who was receiving cimetidine, refused endoscopy. These last two patients had no evidence of gastrointestinal hemorrhage (persistently negative results of stool guaiac tests), and they did not require blood transfusions throughout their hospitalizations.

Nosocomial Pneumonia

Nosocomial pneumonia occurred in 6% of control patients, in 12% of patients receiving sucralfate (relative risk, 2.0; CI, 0.79 to 5.01; P = 0.14), and in 13% of patients receiving cimetidine (relative risk, 2.2; CI, 0.88 to 5.33; P = 0.09). Table 6 shows the distribution of risk factors associated with nosocomial pneumonia. There were no significant differences among the three treatment groups for any of the factors. Nosocomial pneumonia occurred after a mean of 6.9 ±7.2 days in the ICU (median, 9 days). There were no significant differences among the three groups, either by log-rank test or by Cox regression when adjusting for any of the factors in Table 6.

Patients in whom nosocomial pneumonia developed had a mean increase in ICU stay of 18 days compared with those without pneumonia (23.8 ±17.0 days and 5.5 ±6.0 days, respectively; P < 0.001). The median ICU durations of stay were 17 and 3 days, respectively. This increased the mean total duration of hospital stay for patients with nosocomial pneumonia (34.3 ±21.0 days and 14.2 ±15.5 days). The median lengths of hospital stay were 29 and 10 days, respectively. The mortality rate for those in whom nosocomial pneumonia developed was significantly greater than the mortality rate in those without pneumonia (35.5% compared with 19.7%; relative risk, 1.8; CI, 1.06 to 3.07; P = 0.04).

Mortality Rates and Hospital Stays

Prophylaxis did not reduce mortality rate, duration of ICU stay, or total hospital stay (Table 7). However, given our sample size and the observed confidence intervals, we cannot absolutely exclude a benefit of prophylaxis. The duration of ICU stay increased by a mean of 5 days (median, 2 days) in patients with stress-related hemorrhage. This outcome was not influenced by prophylaxis (P = 0.72). The increased length of stay in the ICU was not the result of the hemorrhage. Rather, the patient's other medical conditions were responsible for the prolonged ICU stay.

Drug Safety

Possible adverse drug reactions requiring discontinuation of therapy were seen only in the patients receiving cimetidine: mild confusion in three, somnolence in two, and thrombocytopenia in one. In all instances, cimetidine therapy was discontinued and symptoms resolved within the next 24 hours. Three of the patients were rechallenged with the same dose of cimetidine within 48 hours of drug discontinuation. They tolerated drug resumption without problems and received cimetidine for the duration of their ICU stay. The Adverse Drug Reaction Score ranged from 0 to 3, suggesting that cimetidine probably did not cause the problem [26]. None of the patients bled in whom cimetidine therapy was discontinued.

Discussion

Top Methods Results Discussion Author & Article Info References

Our study shows that prophylaxis with cimetidine or sucralfate did not change the incidence, severity, or outcome of stress-related gastrointestinal hemorrhage. The incidence of clinically important stress-related hemorrhage in patients in the ICU who received prophylaxis with cimetidine (5%) or sucralfate (5%) was not significantly different from that observed in patients not receiving prophylaxis (6%) (P = 0.75). The transfusion requirements were unaffected by prophylaxis. In all groups, the hemorrhage was either limited or occurred as a terminal event. Although patients with stress-related hemorrhage had increased mortality rates compared with patients without hemorrhage, bleeding was not the cause of death. Rather, hemorrhage reflected gastrointestinal involvement in the spectrum of multiple-organ dysfunction syndrome [33-35].

Although our conclusions regarding the lack of efficacy of stress-related gastritis prophylaxis correlate with those of some studies [12, 23, 36], they differ from the results of others [22, 37-39] and of two meta-analyses [40, 41]. There are several potential explanations for these differences. Many of the frequently referenced studies suggesting a benefit of prophylaxis are now more than 10 years old. Because of improvements in ICU care, the conclusions of these older studies may no longer be applicable. Current literature suggests that both the incidence and severity of hemorrhagic stress-related gastritis have decreased independent of the use of prophylaxis [13]. Factors thought to contribute to the decreasing incidence of stress-related gastritis include more aggressive shock management, improved methods of ventilatory support, careful attention to acid-base balance, early enteral or parenteral nutrition, and improved methods to control bleeding diatheses [8, 9, 33, 35, 36, 40, 42, 43]. Our use of a concurrent control group that did not receive prophylaxis allowed us to determine the incidence of stress-related hemorrhage in patients receiving the most current medical ICU care.

A second explanation for the different conclusions may be related to patient selection. Many of the studies investigated a heterogeneous patient group. Patients with trauma, postoperative complications, increased intracranial pressure, extensive burns, or multiple medical conditions were frequently included in the same study. Such different reasons for ICU admission may be associated with different degrees of risk for stress-related bleeding. The conclusions drawn from the total population may not be applicable to each of its constituent groups. We prospectively studied only patients admitted to medical ICUs, a more homogeneous patient population than that of other studies.

A third explanation may be the different criteria for stress-related hemorrhage. Some investigators considered the presence of occult blood in gastric aspirates sufficient evidence of hemorrhage [1, 4], whereas others required bright red blood in the nasogastric tube or a decrease in hematocrit [39]. We used a definition of hemorrhage that suggested clinically substantial blood loss. We did not consider occult bleeding or overt hemorrhage without a decrease in hematocrit as evidence for substantial gastrointestinal hemorrhage.

Finally, most investigators did not use endoscopy to verify that stress-related gastritis was responsible for the hemorrhage [23]. Because patients in the ICU may bleed from many causes, endoscopy is essential to determine whether stress-related gastritis caused the hemorrhage [8]. We verified endoscopically the cause of the hemorrhage. This is important because other lesions may cause bleeding (Table 5). Further, nasogastric tube-induced trauma or coagulopathy can cause hemorrhage or exacerbate minor bleeding due to other causes [42].

The low incidence of hemorrhage in the control group and the failure to show a benefit for prophylaxis was not due to the degree of illness in the patients we studied. The mean APACHE II score of the three groups ranged from 16 to 18. One third of the patients in each group had APACHE II scores greater than 20. In nonsurgical patients, these scores are associated with a death rate of at least 20% [17]. The hospital mortality rate in our patients was 21%. Additional evidence of the severity of illness was the need for mechanical ventilation in two thirds to three quarters of our patients. Finally, 90% of patients had at least one risk factor associated with an increased risk for stress-related hemorrhage [6, 23, 24]. Approximately one third of our patients had three or more of the factors proposed to increase the risk for stress-related ulceration.

Our data showed that only respiratory failure and high-dose corticosteroid administration were independently associated with an increased risk for stress-related hemorrhage. We re-examined the utility of prophylaxis using only patients believed to be at high risk (that is, with either of the previously noted risk factors, APACHE II scores greater than 20.0 or coagulopathy). These analyses also failed to show a benefit for prophylaxis, a finding similar to that of other studies of patients in the ICU at high risk [12, 36, 44].

Although we could not show an appreciable benefit for prophylaxis, we cannot exclude greater reductions in bleeding rates because of our sample size. For example, to exclude a 50% reduction in bleeding rates with prophylaxis would require randomizing 814 patients per study group. Given the decreasing incidence of severe hemorrhage from stress-related gastrointestinal ulceration [13], future studies should be of adequate sample size to detect clinically meaningful benefits of prophylaxis.

Our study suggested an increased incidence of nosocomial pneumonia with cimetidine or sucralfate prophylaxis, although the difference from the control group was not significant. We observed that nosocomial pneumonia developed in 6% of controls, in 12% of those receiving sucralfate, and in 13% of those given cimetidine. It is difficult to compare our results with those reported by others because of large variations in the reported incidences of nosocomial pneumonia. In part, the differences among studies may be related to the use of mechanical ventilation or to the patients studied [15, 45]. There are few other studies for comparison because, until recently, investigators were reluctant to include "no treatment" groups because of concerns regarding stress-related hemorrhage. We are unaware of other studies that have concurrently compared the rates of nosocomial pneumonia for cimetidine, sucralfate, and "no prophylaxis" groups.

The reason for the increased incidence of pneumonia in both the cimetidine and sucralfate groups is unclear. Several studies have suggested that reduction of gastric acid secretion by H₂-receptor antagonists may increase the risk for developing nosocomial pneumonia in patients requiring mechanical ventilation [7, 14, 15], although this conclusion is not unanimous [39]. Gastric pH was greater than 4.0 considerably more often in the cimetidine group than in the other groups. However, this result would not explain the incidence of pneumonia in the sucralfate group, which had gastric pH values similar to those seen in the control group.

Factors proposed to increase the incidence of nosocomial pneumonia include use of mechanical ventilation, duration of mechanical ventilation, presence of nasogastric tube, and enteral feeding [15, 46]. The distribution of these factors among our three groups was not statistically different, suggesting that these conditions alone were not responsible for the doubling of the pneumonia rates.

This issue of prophylaxis-related pneumonia requires further investigation if stress-related gastritis prophylaxis is to be considered. Future studies should be of adequate sample size and should use protected brush specimen techniques or bronchoalveolar lavage to improve the diagnostic accuracy for nosocomial pneumonia [45].

To justify the use of stress-related gastritis prophylaxis for patients entering the ICU, the therapy should be effective and safe. We could not show that either cimetidine or sucralfate prophylaxis affected the already low incidence of stress-related hemorrhage in the ICU. Further, our data suggested that the medications may increase the risk for nosocomial pneumonia. Based on our observations, routine prophylaxis for stress-related bleeding with these agents is not warranted for patients entering the medical ICU.