Trial Design and Oversight
We performed screening and randomization between January 4, 2016, and October 22, 2017, in 33 ICUs in Denmark, Finland, the Netherlands, Norway, Switzerland, and the United Kingdom after obtaining institutional approval at each site. We obtained written informed consent from the patients or their legal surrogates in accordance with national regulations. In most institutions, if the patient or legal guardian was unable to give consent initially, enrollment of patients was allowed on an emergency basis (e.g., with consent from a doctor who was independent of the trial), followed by consent from relatives and the patient to continue participation. If consent was withdrawn or not granted, we asked the patient or surrogate for permission to continue registration of trial data and for permission to include these data in our analyses. The trial protocol and the statistical analysis plan have been published elsewhere and are available with the full text of this article at NEJM.org.13,14 All the authors vouch for the adherence of the trial to the protocol, the accuracy and completeness of the data and analyses, and the reporting of adverse events. The members of the steering committee wrote the draft of the manuscript and made the final decision to submit it for publication. The trial was funded by Innovation Fund Denmark, which had no role in the design of the protocol, the trial conduct, or the analyses or reporting of the data. There was no commercial support for the trial.
This trial was a multicenter, stratified, parallel-group, placebo-controlled, blinded clinical trial. Randomization was performed with a centralized, computer-generated allocation sequence stratified according to trial site and the presence or absence of active hematologic cancer. Patients who were admitted to participating ICUs were screened and, if eligible, were randomly assigned in a 1:1 ratio, with the use of permuted blocks of varying sizes, to receive pantoprazole or placebo. The trial-group assignments were concealed from the patients, clinicians, investigators, trial statisticians, and members of the data and safety monitoring committee. The data and safety monitoring committee oversaw the conduct of the trial and the safety of the participants. Interim analyses were planned to be conducted when 1675 and 2500 patients had been followed for 90 days. The latter interim analysis was not performed, because the target number of patients enrolled was completed before the 90-day follow-up of 2500 patients. The trial data were monitored at the sites by external monitors in accordance with the Good Clinical Practice directive of the European Union and centrally by staff from the coordinating center.
We screened patients who were admitted to the intensive care unit (ICU) for an acute condition (i.e., an unplanned admission), were 18 years of age or older, and had at least one risk factor for gastrointestinal bleeding. Patients were excluded if there were medical reasons, if they were receiving ongoing daily acid-suppressing therapy in the ICU, or if informed consent could not be obtained; 1001 patients met two or more exclusion criteria. A total of 52 patients were excluded immediately after randomization and before the first dose of trial agent because they did not meet the inclusion criteria or fulfilled one or more exclusion criteria; 7 patients were excluded after randomization because the patient or the surrogate did not allow the use of their data. The remaining 267 patients (130 in the pantoprazole group and 137 in the placebo group) who withdrew from the trial at their own or their surrogates’ request allowed the use of their data, but 20 patients or surrogates in the pantoprazole group and 20 in the placebo group did not want further data to be registered except for data on mortality, which we obtained from registries. Some of the process variables and some of the secondary outcomes were missing for these 40 patients.
We screened patients 18 years of age or older who were admitted to the ICU for an acute condition (i.e., excluding elective admissions) and had at least one risk factor for clinically important gastrointestinal bleeding, including shock, use of anticoagulant agents, renal-replacement therapy, mechanical ventilation (expected to last >24 hours), any history of liver disease, or any history of or ongoing coagulopathy (the full definitions of the criteria are provided in the Supplementary Appendix, available at NEJM.org4,13). We encouraged the investigators at all the trial sites to systematically complete the screening procedure for all patients who fulfilled the inclusion criteria. Patients were excluded for the reasons shown in Figure 1 and in the Supplementary Appendix.
Enrolled patients received an intravenous injection of pantoprazole (40 mg suspended in 10 ml of 0.9% sodium chloride) or matching placebo (suspended in 10 ml of 0.9% sodium chloride) (Fig. S1 in the Supplementary Appendix) as a single bolus once daily from randomization until ICU discharge or death, for a maximum of 90 days. If a patient was readmitted to a trial ICU within 90 days, the originally assigned trial regimen was resumed. All other interventions were chosen at the discretion of the clinicians.
The primary outcome was death by 90 days after randomization. The secondary outcomes were clinically important events in the ICU (defined as clinically important gastrointestinal bleeding,4 new-onset pneumonia,15 C. difficile infection, or acute myocardial ischemia; see the Supplementary Appendix); clinically important gastrointestinal bleeding in the ICU (defined as overt gastrointestinal bleeding and at least one of the following four features within 24 hours of gastrointestinal bleeding, in the absence of other causes, in the ICU: a spontaneous decrease in systolic blood pressure, mean arterial pressure, or diastolic blood pressure of 20 mm Hg or more; initiation of treatment with a vasopressor or a 20% increase in vasopressor dose; a decrease in hemoglobin of at least 2 g per deciliter [1.24 mmol per liter]; or transfusion of two or more units of packed red cells); infectious adverse events in the ICU (new-onset pneumonia or C. difficile infection); serious adverse reactions in the ICU (see the Supplementary Appendix); and the percentage of days alive without the use of life support (mechanical ventilation, circulatory support, or renal-replacement therapy; see the Supplementary Appendix) within the 90-day period. Data for outcome measures were obtained from the patients’ files by the trial investigators or their delegates. For 90-day mortality, regional and national registries or direct contact with participants or surrogates were used if available.
We estimated that 3350 patients would be required for the trial to have 90% power to detect a between-group difference of 5 percentage points in 90-day mortality, corresponding to a 20% difference in relative risk at a two-sided alpha level of 5%, under the assumption of a baseline 90-day mortality of 25%.4,13,14 The statistical analysis was performed in accordance with the International Conference on Harmonisation tripartite guideline (Guideline for Good Clinical Practice16) and the statistical analysis plan, by an independent statistician who was unaware of the trial-group assignments.14 We conducted the primary analysis in the intention-to-treat population, defined as all patients who underwent randomization except for the 59 patients who were excluded after randomization; 7 patients did not consent to the use of their data, and 52 patients were excluded immediately after randomization before the first dose of trial medication because they did not meet the inclusion criteria or fulfilled one or more exclusion criteria.17 In the per-protocol population, we excluded patients with one or more major protocol violations (Table S2 in the Supplementary Appendix).16
In the primary analyses, we evaluated data in the two groups using a binary logistic regression adjusted for trial site and active hematologic cancer in the intention-to-treat population18; relative risks with 95% confidence intervals were computed as described in the statistical analysis plan (see the Supplementary Appendix). We evaluated the primary outcome in the per-protocol population and in prespecified subgroups defined by the presence or absence of any history of liver disease, the presence or absence of any history of coagulopathy or ongoing coagulopathy, the type of ICU admission (medical vs. surgical), the presence or absence of shock, the use or absence of use of mechanical ventilation, and a Simplified Acute Physiology Score (SAPS) II19 above 53 or 53 or lower at baseline. The SAPS II is calculated from 17 variables and has a total range from 0 to 163, with higher scores indicating greater severity of disease; a score of 53 was chosen as predicting a 50% mortality rate in the original model.
For secondary analyses, we evaluated all dichotomous outcomes in binary logistic regression analyses of the intention-to-treat population adjusted for stratification variables and predefined risk factors at baseline (age, type of admission [medical, elective surgery, or emergency surgery], and the Sepsis-related Organ Failure Assessment [SOFA] score assessed in the 24 hours before randomization).20 The SOFA score is made up of subscores ranging from 0 to 4 for each of six organ systems (cerebral, circulatory, pulmonary, hepatic, renal, and coagulative), resulting in an aggregated score that ranges from 0 to 24, with higher scores indicating more severe organ failure. We also performed unadjusted chi-square testing for the binary outcome measures. We analyzed percentages of days alive without life support in the 90-day period with the van Elteren test (with adjustment for site only), because the assumptions for Poisson or negative binomial distributions were not met.21 There was no adjustment for multiple comparisons of secondary outcomes, and the results are reported as point estimates with unadjusted 95% confidence intervals. There was no imputation for missing data (details are provided in the Supplementary Appendix). A total of 9 patients were lost to 90-day follow-up. We performed analyses using SAS software, version 9.4 (SAS Institute), and R software, version 3.4.3 (R Foundation), and considered a two-sided P value of less than 0.05 to indicate statistical significance for the primary outcome with 95% confidence intervals.