Clinical Trial Details
— Status: Completed
Administrative data
NCT number |
NCT04826666 |
Other study ID # |
20.436 |
Secondary ID |
|
Status |
Completed |
Phase |
|
First received |
|
Last updated |
|
Start date |
March 31, 2021 |
Est. completion date |
November 4, 2021 |
Study information
Verified date |
March 2021 |
Source |
Centre hospitalier de l'Université de Montréal (CHUM) |
Contact |
n/a |
Is FDA regulated |
No |
Health authority |
|
Study type |
Observational
|
Clinical Trial Summary
Liver transplantation are surgeries associated with important bleeding and often require
perioperative red blood cell (RBC) transfusions. Overall, between 20 and 85 % of liver
transplant recipients receive at least one RBC transfusion during their surgery. Such
transfusions are consistently associated with higher morbidity and mortality, although this
causal association is still under debate in many surgical populations. Despite the lack of
clear causative association between perioperative transfusions and worse outcomes, minimizing
bleeding and transfusions is believed to improve postoperative outcomes. Many perioperative
variables are associated with higher blood loss and need for perioperative transfusions:
liver disease severity, preoperative anemia and coagulopathy, higher cardiac filling
pressures and higher fluid administration, among others. However, few perioperative
interventions have been shown to reduce bleeding and transfusion requirements in this
population. Among them, the use of intraoperative phlebotomies to reduce portal and hepatic
venous pressure during the dissection phase is a promising one, also described in liver
resection surgery.
To further investigate the effects of intraoperative phlebotomies on intraoperative bleeding,
perioperative transfusions and mortality, the Principal Investigator will conduct a
retrospective cohort study with a propensity score based causal analysis.
Description:
Methods
Objective: To investigate the effects of intraoperative phlebotomies on intraoperative
bleeding, perioperative transfusions and mortality among patients who underwent a liver
transplantation.
Study design and participants:
All successive adult patients who underwent a liver transplantation between July 2008 and
January 2021 at the Centre hospitalier de l'Université de Montréal (CHUM) will be included in
the study. Because phlebotomies are mostly performed in patients with a near normal renal
function, patients who were under renal replacement therapy prior to surgery as well as those
who had a glomerular filtration rate below 30 mL/h (based on MDRD formula) will be excluded.
Exposure:
The exposure of interest will be the performance of an intraoperative phlebotomies.
Intraoperative phlebotomies are performed at the beginning of liver transplant surgery to
reduce portal and hepatic venous pressure and thus reduce bleeding and the need for
transfusions. After graft reperfusion, retrieved blood is reinjected to patients to optimize
volemia and cardiac output. Intraoperative phlebotomy is thus a manipulable well-defined
exposure amenable to causal analysis by consistency.
Covariables:
Many variables are associated with bleeding in liver transplantation. Most of them are also
associated with our exposure of interest and are cofounding factors. Phlebotomies will be
more often performed in non-anemic cirrhotic patients with high portal and central venous
pressure, but less often in patients with severe acute disease with end-organ damage such as
renal failure. Thus, patients who received a phlebotomy have different baseline prognostic
characteristics than those who do not receive as phlebotomy. To control for such confounding,
a sufficient set of variables based on a directed acyclic graph (DAG) constructed using
published data and knowledge of the clinical practice (see figure 1) will be included. Since
MELD score is a very robust marker of liver disease severity, it will increase in most
situations of worsening liver disease (such as acute-n-chronic liver failure) and adjust for
all such situations. Since an observational study from the CHUM suggested that intraoperative
bleeding and transfusions have increased since recipients are prioritized by the MELD score,
the calendar year as a covariable will be added, although it is not expected that phlebotomy
practices have significantly changed over time.
Data management:
Data for patients who received their transplantation between 2008 and 2017 is already
available in a dataset used for previous analyses (CHUM Research Ethic Board (REB) approval
#17.036). Data from patients who received a transplantation between January 2018 and January
2021 will be extracted from patients' chart after REB approval or from the dataset used for
another already approved study (CHUM REB #17.251). Data from all these patients will be
merged in a common dataset, cleaned and analyzed.
Data analyses:
Main analyses:
The patient cohort based on the exposure category will be described. Frequencies and
proportions for categorical variables and mean with standard deviation for continuous
variables (or median with quartiles for skewed distributions) will be summarized. Crude
outcome incidence without any relative risk (table 2) will also be presented. For analytical
purposes, intraoperative bleeding will be used as a continuous variable and because many
patients do not receive any perioperative RBC transfusions, perioperative RBC transfusions
between "no transfusion" and "any transfusion" will be dichotomized.
A causal analysis will be conducted based on a balancing score, the propensity score. No
previous sample size calculation was computed as a convenience sample of all transplanted
patients that meet the inclusion criteria will be used. By excluding patients with at least
moderate renal failure prior to surgery (almost only untreated patients), a cohort with more
treated than untreated patients will be obtained. Treated patients are also the ones for whom
clinicians believed a phlebotomy was helpful based on measured covariables (and potential
unmeasured ones ("confounding by indication bias")). Also, many untreated patients may not be
at risk of receiving the intervention (positivity) and not overlap treated patients. Thus, an
average treatment effect in the treated (ATT) by the inverse probability of treatment
weighting (IPTW) will be estimated.
To do so, a propensity score (π_i) for the exposure (intraoperative phlebotomies) based on
all identified and measured previously mentioned confounders will be computed. Quadratic
terms for continuous variables (for more flexibility) and an interaction term between the
MELD score and the central venous pressure (since more severe disease usually have higher
cardiac filling pressure), important drivers of the exposure, will be included. The overlap
of the propensity score between the treated patients and their untreated counterparts will be
evaluated. In case many treated patients do not overlap with any untreated ones, the
specifications of the propensity score model further (remove quadratic terms for example)
will be modified to further restrict the population of interest. The calculated propensity
score to compute weights will be used and create an untreated pseudo-population comparable to
the treated ones (conditional exchangeability); Weights for treated patients and weights will
be used for untreated patients. In case extreme weighs are estimated, truncation (between 1%
and 5% percentiles of the propensity score distribution depending on overlap effect) will be
used to minimize variance and effect of near violations of practical positivity. The
population of interest may be further restriced, if deemed necessary. The pseudo-population
will be described using weighted descriptive statistics and the balancing effect of the
weights will be verified. The causal marginal effect on bleeding will be estimated using a
weighted mean difference and the causal marginal effect of transfusions using a weighted risk
difference. Survival up to 1 year will be reported using a marginal structural model using a
weighted proportional hazard Cox model and express a causal marginal hazard ratio. Results
will be expressed with non-parametric bootstrap percentile 95% confidence intervals.
Senstivity analyses:
IPTW analyses may be more sensitive to misspecification of the propensity score model as well
as have a higher estimated variance. Thus, a sensitivity analysis to estimate ATT for all
outcomes will be conducted using a propensity score based matching analysis. This analysis
will use a 1:2 (1 treated and 2 untreated) greedy matching using a caliper (0.2 linear
propensity score standard deviation). For matching, balancing between groups by comparing
covariables' central tendency measurement and variance will be explored, including quadratic
terms for continuous variables, as well as q-q plots. Distribution homogeneity between groups
will also be assessed using Kolmogorov-Smirnov tests. The matching specifications (caliper,
matching ratio) may bemodified if group balancing is not satisfactory. However, the
propensity score specifications will not be modified, to maintain consistency across
analyses. Once balancing is considered appropriate to estimate an ATT, the causal effects
will be estimated by using the Abadie-Imbens estimator. The Abadie-Imbens variance will be
used to compute 95% confidence intervals. R software (R Core Team, 2020, version 4.0.3) will
be used, as well as the Matching, survival, survey and tableone packages. IPTW analyses and
bootstrap will be computed manually.
Subgroup analysis:
The effect of phlebotomies is considered to be mechanistically mediated by reducing portal
pressure and splanchnic congestion. The effect of the intervention should thus be stronger in
patients with cirrhosis. Thus, a subgroup analysis will be conducted by conduction the
primary analysis (IPTW and bleeding) only in patients transplanted for cirrhosis by excluding
retransplantations and transplantation for acute liver failure.