Page 27 - Physical activity in recipients of solid organ transplantation - Edwin J. van Adrichem

Physical activity in recipients of solid organ transplantation
Endpoints
Cardiovascular and all-cause mortality were the primary endpoints of this study. With a
continuous surveillance system, an adequate and up-to-date collection of patients’ data for
events and mortality could be ensured. CV and all-cause mortality were recorded until May
2009 without loss of follow-up. Cardiovascular death was de
Statistical analysis

ned as deaths in which the
principal cause of death was cardiovascular in nature (ICD-410-447).
Data are presented as means and standard deviations (SD) or median and interquartile
ranges (IQR) for skewed distributed variables. Proportions are shown for categorical data.

e levels of cardiac markers were initially analyzed per tertile of PA level (strati
gender). A linear trend analysis was performed on the di

erences in cardiac markers

ed for
between groups. Log-transformation was applied for variables with a skewed distribution,
and hazard ratios are reported with 95% con

dence intervals (CI).
To investigate whether the cardiac markers were independently associated with CV
and the all-cause mortality multivariable, Cox-regression analyses were performed and
adjusted for potential confounders (age, gender, eGFR, and physical activity). For these
analyses, cardiac markers were transformed into z-scores. Potential interaction e

ects
between the cardiac markers and PA were tested in both crude and adjusted models.
Cox proportional hazards regression models were applied to study the association
between PA and CV- and all-cause mortality. Assumptions for survival analyses were
tested by interactions with a time covariate. Multivariate Cox regression analyses were
adjusted for age, gender (Model 2), history of cardiovascular events (Model 3), eGFR,
urinary protein excretion (Model 4), and MR-pro-ANP, NT-pro BNP, and Hs-TnT
(Model 5).
Additional analyses were performed to examine the potential mediation e
cardiac ischemia and heart failure on the association between PA and CV-mortality.
pathways by which cardiac markers could potentially mediate the e
CV-mortality, respectively.
corrected bootstrap con
mediated was calculated by dividing the coe
signi

dence intervals with 2000 repetitions. Finally, the proportion
cance of mediation was considered present with
lower and upper bound of the 95% con
cient of the indirect e
p
ect was tested with bias

ect of
e

ect of the level of PA
on cardiovascular mortality are shown in Figure 1.

e pathways were assessed with
mediation analysis utilizing the Preacher and Hayes method.

e total e

CV-mortality was estimated using regression analysis.

28,29
e indirect e

ect of PA on
ect of PA on CV-
mortality, through cardiac damage, was calculated by computing the product of two
coe

cients obtained with a regression analysis of the cardiac marker with PA and with

e signi

cance of the indirect e

ect by the total e
<0.05 if zero was not between the

ect.

e mediation analyses were additionally adjusted for gender, age, and eGFR.
e

dence interval of the indirect e

ect. Two-sided
signi

cance tests were used (
alpha
<0.05), and analyses were performed with IBM SPSS
statistical software version 23.0 (IBM SPSS, NY, USA) and STATA/SE 14.0 (StataCorp,
TX, USA).
25.
2