Primary analysis
The main analyses will compare those allocated TXA with those allocated placebo on a modified intention-to-treat basis, excluding patients who received neither dose of the allocated trial treatment. We will present the results as effect estimates (relative risks) with a measure of precision (95% confidence intervals) (see Additional file 1: Table S2). Additionally, we will present results of the primary analysis adjusted for all baseline covariates. If baseline covariates are associated with the outcome, adjusting for any chance imbalances in baseline risk will increase statistical power. We will not present risk differences because they are not a generalisable measure of the treatment effect and are dependent on baseline risk. The effect of TXA will also be examined graphically using cumulative incidence curves presented with their associated hazard ratios and log-rank p values (see Additional file 1: Figure S2) [49]. The effects of TXA on death due to bleeding in the HALT-IT trial will be set in the context of other trials of TXA for acute severe haemorrhage (the CRASH-2 and WOMAN trials).
Primary outcome
Death due to bleeding within 5 days of randomisation is the primary outcome. As outlined in the section ‘Change in primary outcome’ above, cause of death is assigned by local investigators who provide a narrative of the events leading to death. The cause of death narratives are reviewed by the principal investigator (who is blind to treatment allocation) and queried if more information is needed to confirm whether death is due to bleeding or another cause. Furthermore, due to double-blind nature of the trial, the coding of the cause of death cannot be affected by the patient’s randomised group. For more details, please see accompanying information in the section ‘Change of primary outcome’.
Secondary outcomes
We will assess the effect of TXA on the following secondary outcomes. Unadjusted analyses will be presented in the main text and although we do not expect any baseline imbalances, to complement the unadjusted analyses and potentially increase statistical power (if covariates are associated with the outcome) we will present results of the analyses adjusted for all baseline covariates in an appendix.
Rebleeding
Rebleeding generally occurs in approximately 10–25% of patients with acute GI haemorrhage and is associated with an increased risk of death due to bleeding [50]. A clinical diagnosis of rebleeding is made by the treating clinician based on the presence of any of the following criteria, as defined in a data collection guide. These criteria for rebleeding were recommended by a methodological framework for trials in GI bleeding following an international consensus conference [51]:
Haematemesis or bloody nasogastric aspirate > 6 h after endoscopy
Melaena after normalisation of stool colour
Haematochezia after normalisation of stool colour or after melaena
Development of tachycardia (HR > 110 beats per min) or hypotension (SBP ≤ 90 mmHg) after ≥ 1 h of haemodynamic stability (i.e. no tachycardia or hypotension) in the absence of an alternative explanation for haemodynamic instability such as sepsis, cardiogenic shock, or medication
Haemoglobin drop of > 2 g/dl after two consecutive stable values (< 0.5 g/dl decrease) ≥3 h apart
Tachycardia or hypotension that does not resolve within 8 h after index endoscopy despite appropriate resuscitation (in the absence of an alternative explanation) associated with persistent melaena or haematochezia
Persistently dropping haemoglobin of > 3 g/dl in 24 h associated with persistent melaena or haematochezia
It should be noted that patients may continue to have haemodynamic instability, falling haemoglobin levels or persistent melaena or rectal bleeding for hours and even days after bleeding has stopped, making these patients difficult to categorise; however, these criteria are more likely to indicate rebleeding than equilibration [51].
Most rebleeding tends to occur within 5 days of the index bleed [35,36,37]. We believe that TXA will be most effective at reducing the risk of rebleeding soon after the index bleed when blood plasma concentrations of the drug are above the level needed to inhibit fibrinolysis [52]. To assess whether TXA reduces rebleeding, we will analyse the effect on early rebleeding within 5 days of randomisation (see Additional file 1: Table S2).
Although rebleeding is most common within the first 5 days after the index bleed, TXA will have been metabolised within about 2 days of randomisation, with the blood plasma concentration falling below the level needed to inhibit fibrinolysis within around 24 h. As such, we will examine the effect on rebleeding within 24 h of randomisation. We hypothesise that TXA will be less effective for late rebleeding occurring days or weeks after the drug has been eliminated. To investigate this we will assess the effect of TXA on rebleeding within 28 days (see Additional file 1: Table S2). If our hypothesis is correct, the inclusion of late rebleeding events should dilute the treatment effect.
Death due to bleeding within 24 h and 28 days
Tranexamic acid will be eliminated within about 2 days of randomisation, with blood plasma levels falling below those needed to inhibit fibrinolysis within around 24 h. Furthermore, patients with acute GI haemorrhage bleed to death quickly, with many deaths due to bleeding occurring within the first day. Evidence from other trials suggests that this is where the greatest treatment benefit will be observed. As such, we will analyse the effect of TXA on deaths due to bleeding within 24 h of randomisation. Conversely, because there may be a weaker treatment effect on late deaths due to bleeding that occur several days or weeks after randomisation, we will also analyse the effect on death due to bleeding within 28 days of randomisation (see Additional file 1: Table S2). We expect to observe a smaller treatment effect when including late bleeding deaths due to dilution towards the null.
Mortality
We will analyse the effect of TXA on all-cause and cause-specific mortality at 28 days. Specific causes of death to be analysed include death due to bleeding, thrombosis, organ failure, pneumonia, sepsis, malignancy and other causes (see Additional file 1: Table S3). We will also examine the temporal distribution of causes of death by days since randomisation using a frequency bar chart (see Additional file 1: Figure S3). Based on its mechanism of action and data from large randomised trials, we do not expect TXA to reduce deaths from non-bleeding causes like cancer or sepsis or to reduce late deaths from bleeding.
Endoscopic, radiological and surgical procedures for GI bleeding
We will assess the effect of TXA on diagnostic and therapeutic endoscopic and radiological procedures and surgical interventions (see Additional file 1: Table S5). It remains unclear whether TXA reduces the need for surgery in GI bleeding [41]. In large trials of TXA for postpartum and traumatic haemorrhage, there was no evidence of an effect on surgical interventions except for laparotomy for bleeding [39, 40]. If TXA reduces GI bleeding, it has the potential to reduce the need for some endoscopic, radiological and surgical procedures. While we do not expect TXA to influence diagnostic endoscopic and radiological procedures planned around the time of hospital admission and randomisation, there is potential to reduce the need for diagnostic procedures planned after resuscitation, and, therefore, after randomisation [43]. Similarly, therapeutic procedures and surgical interventions planned and undertaken after diagnosis also have the potential to be influenced by TXA. It is not possible to look at procedures by time as this information was not recorded, and although type of procedure can be used as a rough indication of timing, therapeutic or surgical procedures planned around the time of randomisation could still dilute the effect estimates towards the null.
Blood transfusion
Since blood transfusion is mostly determined by blood loss prior to randomisation, we do not expect to see a marked reduction in the need for blood transfusion with the use of TXA [43]. Major haemorrhage protocols dictate the type and volume of blood components that patients receive based on presenting clinical signs such as blood pressure and estimated blood loss. Furthermore, survivor bias could lead to higher transfusion rates in the TXA group. In keeping with this, a systematic review of TXA for GI bleeding found no reduction in transfusion [41]. Although TXA has the potential to reduce transfusion for blood lost after randomisation, e.g. after rebleeding, we did not collect data on date and time of transfusion. Any effect on late transfusions is likely to be obscured by early transfusions for blood lost pre-randomisation. We will assess the effect of TXA on the use of whole blood or packed red cells, frozen plasma and platelets comparing the frequency of transfusion and the mean number of (adult-equivalent) units transfused (see Additional file 1: Table S5).
Thromboembolic events
An individual patient data meta-analysis of the WOMAN and CRASH-2 trials found evidence of a reduction in myocardial infarction with TXA (OR = 0·64, 95% CI 0·43–0·97; p = 0·037) and no evidence of an increased risk of fatal vascular occlusive events (OR 0·73, 95% CI 0·49–1·09; p = 0·120) or other non-fatal events [53]. While this finding is reassuring, we cannot exclude the possibility of some increased risk with TXA, particularly as patients with GI bleeding are older than those with traumatic or postpartum haemorrhage and many have multiple comorbidities. Older age is associated with a pro-coagulation haemostatic profile including elevated fibrinogen and plasminogen activator inhibitor 1 and reduced clotting time [54,55,56]. A systematic review of TXA for the treatment of upper GI bleeding found no evidence for a difference in the risk of thromboembolic events but lacked power [41]. We will examine the effect of TXA on fatal and non-fatal pulmonary embolism, deep vein thrombosis, stroke and myocardial infarction (see Additional file 1: Table S6).
Complications
We will analyse the effect of TXA on renal, hepatic and respiratory failure, cardiac events, sepsis, pneumonia and seizures (see Additional file 1: Table S6). If TXA reduces death due to bleeding, patients in the tranexamic group will survive for longer on average and may, therefore, be at greater risk of complications such as sepsis, pneumonia and organ failure. Generally, death due to bleeding tends to occur soon after bleeding onset whereas infections and organ failure take several days to occur. On the other hand, if TXA reduces bleeding it may reduce liver failure because bleeding can lead to the deterioration of liver function. Although there is evidence that high-dose TXA can cause seizures, we do not expect to see an increase in seizures with the low dose given in the trial.
Self-care capacity
Patients self-care capacity will be measured using the Katz Index of Independence in Activities of Daily Living (Katz ADL) [57]. Participants’ performance in six functions (bathing, dressing, toileting, transferring, continence and feeding) is assessed at the time of discharge from the randomising hospital or in hospital 28 days after randomisation. A score of 1 is assigned to each function the individual can perform independently and they are summed to produce a total score. A score of 6 suggests full function, 4 suggests moderate impairment, and 2 or less suggests severe functional impairment. We expect that reduced blood loss in patients who receive TXA will result in less functional impairment. That said, it is possible that patients in the treatment group will be discharged faster which could mask improvements in self-care capacity at the time of discharge. To assess this hypothesis we will compare the difference in mean Katz ADL score in survivors in the TXA and placebo groups as well as the proportion of patients with no impairment (6), mild to moderate impairment (3–5) or severe impairment (0–2), (see Additional file 1: Table S6).
Days spent in the intensive care or high-dependency unit
We will analyse the effect of TXA on number of days spent in the intensive care unit (ICU) or high-dependency unit (HDU). We will compare the difference in mean number of days spent in the ICU or HDU in the TXA and placebo groups (see Additional file 1: Table S6). Because beds in these units can be limited, we may not see an effect on this outcome measure.
Adverse events
Data on the number of adverse events (AEs), serious adverse events (SAEs) and suspected unexpected serious adverse reactions (SUSARs) reported up to 28 days after randomisation will be presented. We will present a summary table in an Additional file 1 to describe the type of AE, Medical Dictionary for Regulatory Activities (MedDRA) preferred term (PT), MedDRA system organ class (SOC) and the number of occurrences and outcomes (completely recovered, recovered with sequelae, or died) in the TXA and placebo groups. With events grouped by MedDRA SOC, we will compare the frequency of events between trial arms using an unadjusted modified Poisson regression model (see Additional file 1: Table S7). AEs with evidence that they may be increased by TXA (i.e. seizures and thromboembolic events), will be analysed on an individual basis as well as recurrent episodes of GI bleeding reported as AEs.
Subgroup analyses
We will conduct the following subgroup analyses for the primary outcome of death due to bleeding: time to treatment, location of bleeding, cause of bleeding and clinical Rockall score. We will fit interaction terms with randomised group in a Poisson regression model with robust error variance from the sandwich estimator [58]. Interaction tests (the Wald test) will be used to explore whether the effect of treatment (if any) differs across these subgroups. Results will be presented as unadjusted and adjusted effect estimates with a measure of precision (95% confidence intervals) and p value for the test for interaction (see Additional file 1: Table S4). Except for time to treatment, statistically significant heterogeneity between subgroups is required, as determined by the test for interaction p value, and not just statistical significance of a result in a specific subgroup [59].
Although treatment group is randomised within subgroups, the factors defining subgroups are not randomised. Several baseline characteristics are associated with the subgroup variables. For example, early treatment is correlated with bleed characteristics and patient characteristics (see Fig. 2), some of which confer a higher clinical Rockall score, suggesting that patients with more severe bleeding are treated earlier. Since these factors are also associated with mortality, they could potentially confound the interaction between time to treatment and the treatment effect.
If TXA is shown to be effective and the treatment effect varies by time to treatment, there is potential to intervene on time to treatment in order to increase the treatment effect. Although we cannot intervene on location of bleeding, cause of bleeding or clinical Rockall score, we are interested in ascertaining causal interaction of these factors with the treatment effect rather than simply assessing effect heterogeneity. As such, we will adjust all subgroup analyses for potential confounders [60]. Selection of potential confounders is based upon review of unblinded data within the trial to date in order to identify prognostic baseline characteristics that are associated with the subgroup variables. Potential confounders include age, time to treatment, SBP, HR, signs of shock, location of bleeding, suspected active bleeding, comorbid liver disease and suspected variceal bleeding. Signs of shock may be collinear with HR or blood pressure, and suspected variceal bleeding may be collinear with comorbid liver disease – if so, signs of shock and suspected variceal bleeding will not be included in the models. The final models remain to be determined because the outcome of interest is the treatment effect and the association between the potential confounders and the treatment effect cannot be assessed before unblinding.
Time to treatment (≤ 3 h, > 3 h)
Trials of TXA in traumatic and postpartum haemorrhage provide evidence that early treatment (within 3 h of bleeding onset) confers the most benefit, while late treatment is ineffective [39, 53, 61]. As such, we plan to conduct a subgroup analysis of the treatment effect stratified by time to treatment. Patients with GI bleeding may not experience symptoms immediately so time of symptom onset may not accurately reflect time of bleeding onset. Time to treatment may, therefore, be underestimated. Because few patients are treated early (within 3 h), there may be low power to detect an interaction if one exists. As such, we will analyse time to treatment as both a categorical (≤ 3 h, > 3 h) and continuous variable because the latter will preserve more information so should have more power. However, a limitation of modelling time to treatment as a continuous variable is the need to specify the form of the association. To assess non-linearities, we will fit a logistic regression model and use a likelihood ratio test. Any differences between the two approaches will be noted.
There is strong prior evidence to expect a time-to-treatment interaction, with early treatment conferring a greater benefit and late treatment being ineffective and possible even harmful [53, 61]. As such, for the subgroup analysis of time to treatment we do not require as strong evidence against the null hypothesis of homogeneity as we might usually require. Most trials lack power to detect heterogeneity in treatment effects and the lack of a statistically significant interaction does not mean that the overall treatment effect applies to all patients. Due to prior evidence that early treatment is more effective, we will consider the time to treatment subgroup analysis in the context of the existing data (in particular data from the CRASH-2 and WOMAN trials) on the time-to-treatment interaction and will rely more on scientific judgment than on statistical tests.
Location of bleeding (upper GI, lower GI)
We will examine the effect of TXA on death due to bleeding stratified by location (upper versus lower GI). Evidence suggests the rates of rebleeding and mortality after upper and lower GI bleeding are similar [34], and there is no reason to expect the effect of TXA to vary substantially by location of bleeding in the GI tract. Unless there is strong evidence against the null hypothesis of homogeneity of effects (i.e. p < 0.01), the overall relative risk will be considered the most reliable guide to the approximate treatment effect in all patients.
Suspected variceal bleeding and comorbid liver disease (yes, no/unknown)
Outcomes in acute GI bleeding vary by cause of haemorrhage. Variceal bleeding is associated with the highest risk of rebleeding and death. Oesophageal varices are dilated submucosal veins that usually develop because of portal hypertension, often due to cirrhosis. Haemostasis is disturbed in patients with liver disease because many of the pro- and anti-coagulation factors and components of the fibrinolytic system are produced by hepatic parenchymal cells in the liver, although the overall sum of effects are debated [62,63,64]. Any resulting imbalance in coagulation or fibrinolysis may alter the antifibrinolytic activity of TXA; however, the direction of this potential effect remains to be determined. We will examine the effects of TXA on death due to bleeding in patients with suspected variceal bleeding and comorbid liver disease compared to other or unknown causes of bleeding. Unless there is strong evidence against the null hypothesis of homogeneity of effects (i.e. p < 0.01), the overall relative risk will be considered the most appropriate measure of effect.
Clinical Rockall score (1–2, 3–4, 5–7)
We will assess the effect of TXA stratified by the clinical (pre-endoscopy) Rockall score, a widely used risk scoring system for GI bleeding. The score is derived from age, comorbidities, signs of shock, HR and SBP, all of which are independent predictors of mortality. Although originally developed for upper GI bleeding [17], the Rockall score has also been shown to be predictive of mortality in lower GI bleeding [34]. We do not expect the treatment effect to vary by Rockall score. Unless there is strong evidence of an interaction (p < 0.01), we will present the overall relative risk as the most appropriate measure of effect.
Missing data
Based on the data collected to date, we expect loss to follow-up to be minimal (i.e. less than 1% missing data on the primary outcome). Any missing values will be reported but not imputed.