Plasma Transfusion in Major Vascular Surgery

Overview

BACKGROUND * Major blood loss is frequent in open repair of ruptured and intact abdominal aortic aneurysm (AAA) as well as in aorto-bifurcated prosthesis insertion due to aortoiliac occlusive disease. * Major blood loss is associated with death, post-operative complications and coagulopathy. * Data from randomized trials in trauma patients indicate that a high plasma to red blood cell (RBC) transfusion ratio reduces 30-day mortality. * No randomized trial data are available for the AAA population. * Observational data demonstrate, that a high plasma:RBC transfusion ratio associates to a lower 30 day mortality. However, the reports are based on small cohorts of 78-165 patients, short term outcomes and lack information on major adverse events such as cardiac and respiratory. * The Danish Vascular Registry (DVR), covering 1996-2018, contains data on approx. 4,400 ruptured and 8,200 intact (elective/symptomatic) AAA repairs, and 5,400 open aortoiliac repairs due to occlusive disease. Expected total count 1997-2018: 17,000. * The Danish Transfusion Database (DTDB), covering approx. 1997-2018, contains information on units of RBCs, plasma and platelets transfused. A unique patient identification number (CPR) allows merging of all data set. OBJECTIVE To identify whether resuscitation with a high plasma to RBC ratio associates to improves survival in open abdominal aortic surgery as compared to a low plasma to RBC-ratio. PICO * Population: Open abdominal aortic surgery * Intervention: "High FFP": FFP to RBC unit ratio of 2:3 to 3:3 (0.7 - 1.0) * Comparison: "Low FFP": FFP to RBC unit ratio of 0:3 to 1:3 (0.0 - 0.3) * Outcome: All-cause mortality 90 days following surgery. DATA SOURCES CPR, Danish Civil Registration System. DNPR, Danish National Patient registry. DVR, Danish Vascular registry. DPDB, The Danish national Prescription DataBase.

STATISTICAL ANALYSIS PLAN The primary analysis will be a stratified cox regression model. STRATIFICATION: * type of surgery/condition (ruptured AAA vs intact AAA vs aorto-iliac occlusive disease, source: DVR) * sex (CPR) * Center (Rigshospitalet vs. Gentofte vs. Slagelse vs. Odense vs. Kolding vs. Aarhus vs. Viborg vs. Aalborg, source: DVR) COVARIATE ADJUSTMENT: * calendar time (DVR) * age (CPR) * Carlsons comorbidity index score (DNPR) * Priority (Acute vs. Sub-acute vs. Elective, source: DVR) * Use of anti-thrombotic drugs (DPDB). A covariate of 4 levels (ATC code is noted in parenthesis). 1. None vs. 2. Anti-platelet therapy * acetylsalicylic acid (B01AC06) or * dipyridamole (B01AC07, eg. persantin or asasantin) 3. Anti-platelet therapy "thienopyridines-like drugs" * clopidogrel (B01AC04) or * ticagrelor (B01AC24, eg. Brilique) or * prasugrel (B01AC22, eg. Efient) Comment: patients prescribed thienopyridines in combination with aspirin will be included in this "thienopyridine anti-platelet group 3" 4. Anti-coagulant therapy * Vitamin K antagonists (B01AA, eg. warfarin or phenprocoumon) * Low-molecular weight heparine (B01AB01-10) * Direct thrombin inhibitors (B01AE, eg. Pradaxa/Dabigatran) * Direct factor Xa inhibitors (B01AF, eg. Rivaroxaban/Xarelto). Comment: patients prescribed both anticoagulant- and antiplatelet therapy and will be included in this "Anti-coagulant group 4". ADDITIONAL ANALYSES: 1. Stratify the population into 4 groups according to the total transfusion requirement 1. patients receiving 10 units or less of any blood product \< 24 hrs or 11-15 units \< 48 hrs.vs. 2. patients receiving 11-15 units of any blood product \< 24 hrs. or 16-20 units \< 48 hrs. vs. 3. patients receiving 16-20 units of any blood product \< 24 hrs. or 21-25 units \< 48 hrs. vs. 4. patients receiving more than 20 units of any blood product \< 24 hrs. or more than 25 units \< 48 hrs. 2. Outcome predicted by a joint function (general interaction) of total plasma transfusion and total blood cell transfusion will be assessed in an exploratory way by inspection and by agnostic modelling in the mold of Multivariate Adaptive Regression Splines (MARS) and recursive partitioning, i.e. Classification And Regression Trees (CART). 3. Redefine intervention and control group as 4th and 1st quartile of FFP:RBC ratio. Initially, the population will be divide into 4 groups according to quartiles and compared the population below 1st quartile with the population above the 4th quartile, which will define the low vs. the high FFP group, resp. However, to allow for stratification for operation type (ruptured AAA vs. intact AAA vs. occlusive disease) it may be necessary to adjust the percentile cut to retain power in the analyses. For instance, the population may be cut according to tertiles, or, if there is sufficient data, cut by quintiles (5 groups) or deciles (10 groups). 4. Confine the population to patient with blood loss above 50 % of total blood volume (calculated by Naddler's equation accounting for sex, weight and height). If height and weight are not available, the registered blood loss must exceed 2 L in females and 2.5 L in males. 5. Adjusting exclusively for calendar year, sex, age, Charlson's comorbidity index score, and center (ie, excluding priority and antithrombotic therapy). MISSING DATA Missing data will not be an issue for the number of blood transfusions because units of blood products transfused are used as an inclusion criterion. All remaining covariates are discrete, and missing data for each of those will be included as separate parameters (factor level). STATISTICAL SIGNIFICANCE LEVEL Bonferroni adjustment of the significance level will be applied to control for multiple testing. With one primary and four secondary outcomes, only a P-value below 0.01 (0.05/5) will be considered statistically significant. A P-value between 0.01 and 0.05 will be considered borderline significant.