Heart, Lung and Circulation (2024) 33, 710–720 ORIGINAL ARTICLE

1443-9506/23/$36.00
https://doi.org/10.1016/j.hlc.2023.11.004

Prophylaxis to Prevent Cardiotoxicity in

Patients Receiving Anthracycline for
Breast Cancer: A Combined Bayesian and
Frequentist Network Meta-Analysis of
Randomised Controlled Trials
Prajith Jeyaprakash, BMed, MD, MSc, Med a,b, Sukhman Sangha, MBBS a,b,
Gary Low, MBBS, PhD c,d, Christopher Yu, MBBS a,b,
Faraz Pathan, MBBS, PhD a,b, Kazuaki Negishi, MD, PhD, MSc a,b,*
a
Department of Cardiology, Nepean Hospital, Sydney, NSW, Australia
b
Department of Academic Medicine, Sydney Medical School Nepean, Faculty of Medicine and Health, Charles Perkins Centre Nepean, The
University of Sydney, Sydney, NSW, Australia
c
Department of Research Operations, Nepean Hospital, Sydney, NSW, Australia
d
Professorial Unit, The George Institute for Global Health, University of New South Wales, Sydney, NSW, Australia

Received 10 November 2022; received in revised form 31 October 2023; accepted 3 November 2023; online published-ahead-of-print 5 January 2024

Background The benefits in survivorship gained with anthracycline (ANT)-based chemotherapies for breast cancer
are unfortunately mitigated for some patients by irreversible cardiotoxicity. Randomised controlled
trials (RCTs) have explored multiple cardioprotection options, however, it remains unclear which drug
is most effective in preserving left ventricular ejection fraction (LVEF). This study aimed to perform a
systematic review and network meta-analysis, using Bayesian and frequentist approaches, of RCTs
evaluating cardioprotective agents.
Methods Two authors searched four databases (CENTRAL, Cochrane Reviews, MEDLINE, SCOPUS), to find
RCTs evaluating cardioprotective agents. Trial populations were limited to patients with breast cancer
without prior ANT exposure. The primary outcome was mean LVEF change pre and post ANT dosing.
Our primary analysis utilised a Bayesian approach, while our sensitivity analysis used frequentist
methodology (Prospero registration number CRD42020199580).
Results From 4,007 search results, we identified 12 RCTs, with their various trial arms considered separately—
nine beta-blocker (BB), two angiotensin-converting enzyme inhibitor /angiotensin receptor blockers
[(AA)1BB=AABB], one AA, one spironolactone, one statin—evaluating 1,126 patients (age 50.5 years).
Bayesian network meta-analysis showed no difference in LVEF preservation between AA (1.3%, 95%
credible interval [-0.20, 2.9]), BB (0.77, [-0.21, 1.8]), AABB (0.84 [-1.1, 2.8]), spironolactone (0.72, [-2.3, 3.7])
or statin (0.60, [-2.4, 3.6]) when compared against placebo. However, the frequentist analysis showed
benefits from using AA (mean difference, 1.32% [0.32, 2.33]) and BB (mean difference, 0.76% [0.12, 1.4]).
Conclusions There is insufficient evidence to support prophylactic cardioprotection to prevent EF reduction. How-
ever, frequentist analysis suggested that AA or BBs provide cardioprotection. Thus, for those already on
other anti-hypertensives, switching to AA or BBs could be considered.
Keywords Cardiotoxicity  CTRCD  Cardioprotection  Network meta-analysis  Breast cancer

*Corresponding author at: University of Sydney Nepean Clinical School, Level 5, South Block, PO Box 63, Penrith, NSW 2751, Australia; Email: Kazuaki.Negishi@
sydney.edu.au; X: @kaznegishi
Ó 2023 The Author(s). Published by Elsevier B.V. on behalf of Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS) and the Cardiac
Society of Australia and New Zealand (CSANZ). This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Preventing Anthracycline Cardiotoxicity: A Network Meta-Analysis 711

Introduction Meta-Analyses recommendations. The Medical Subject

Heading terms used included “anthracycline”, “cancer
Anthracyclines have remained the most effective chemo- therapeutics-related cardiac dysfunction (CTRCD)”, car-
therapy against malignancy for 30 years [1]. They work by dioprotective”, “cardiotoxicity” and “RCT”. References from
inhibiting topoisomerase 2 enzymes that are normally relevant systematic reviews were manually searched by two
required to reseal DNA helices for DNA replication [2]. reviewers, and additional records were incorporated. Dis-
While this is of benefit when stopping the rapid proliferation putes in study selection were discussed with and resolved by
of tumour cells and improving cancer survival, the second- a third senior author. The full search strategy performed for
ary cardiotoxic effects can result in cumulative dose-related each database is outlined in Supplementary Table 1.
cardiotoxicity, and debilitating heart failure [3]. Chemo- We defined the study population as adults over the age of
therapy protocols have been modified in recent years to 18 years, with no pre-existing cardiac dysfunction, who were
minimise lifelong anthracycline (ANT) exposure, and due to receive their first dose of ANT-based chemotherapy
guidelines recommend that a cumulative doxorubicin dose for breast cancer. Trials were only included if patients allo-
greater than 200–500 mg/m2 may result in higher rates of cated to the intervention arm were newly prescribed pro-
cardiotoxicity [4]. phylactic drug prior to ANT commencement. Their control
Known risk factors for cancer therapeutics-related cardiac groups received either no prophylactic therapy or placebo
dysfunction include cumulative lifetime dose of ANT, pre- drug. The key outcome of interest was the mean change in
existing cardiac disease, and concurrent use of cardiotoxic LVEF, measured as a percentage of cardiac function by either
agents such as trastuzumab for breast cancer [5]. Despite transthoracic echocardiography or cardiac magnetic reso-
these efforts, left ventricular ejection fraction (LVEF) decline nance imaging. Included articles needed to assess LVEF us-
is seen in 13.6% of patients receiving ANT-based chemo- ing a method endorsed by appropriate Society Guidelines
therapy [6]. [17]. We excluded paediatric trials to minimise heterogeneity
Multiple randomised controlled trials (RCTs) have in the population. Conference and poster abstracts unpub-
demonstrated varying degrees of success with prophylactic lished as full texts were also excluded.
administration of drugs such as beta-blockers (BB),
angiotensin-converting enzyme inhibitors (ACEi) and
angiotensin II receptor blockers (ARB) [7–11]. Furthermore,
Study Characteristics
there is limited data comparing the efficacy of different drug Baseline characteristics were extracted by two independent
classes. As a result, there is no clear consensus on whether reviewers into independent spreadsheets and included
any one cardioprotective drug is beneficial in preserving participant numbers, mean age, sex, ANT dose, imaging
LVEF and this is reflected in the latest guidelines with a Class modality and cardioprotective drug target dose. Extracted
IIa recommendation, encouraging its use [12]. data was re-examined for translational errors on two sepa-
To understand the discrepancies between published RCTs, rate occasions by three reviewers. The standardised mean
several meta-analyses have been performed [13–16]. Unfor- difference in LVEF pre and post ANT was the primary
tunately, these meta-analyses have reported mixed conclu- outcome measure. The standard deviations (SDs) and confi-
sions, resulting in confusion at the highest level of evidence. dence intervals (CIs) of the difference in means were also
Differences have been predominantly driven by dissimilar obtained, and Cochrane Handbook tools were used to
statistical methods and a lack of precision regarding study calculate missing SD data using significance values and CIs
inclusion criteria. when required [18]. Corresponding authors from included
In light of this, we aimed to consolidate available literature RCTs were also contacted to obtain missing statistical data, if
using a comprehensive and methodical statistical approach. further information was required. We utilised conversion
We performed a systematic review and network meta- factors described in the European Society of Cardiology
analysis of RCTs that assessed the efficacy of prophylactic consensus statement to standardise ANT dosing into a single
cardioprotection among patients with breast cancer who measure termed doxorubicin equivalent ANT dose (EAD)
received ANT-based chemotherapy for the first time. [19]. This allowed idarubicin and epirubicin doses to be
scaled to doxorubicin doses with scaling factors of 0.53 and
0.7, respectively.

Methods
Statistical Analysis
Study Selection Statistical analysis was performed by an academic biostatis-
We performed a comprehensive search of randomised con- tician (GL). A Bayesian network meta-analysis was used as
trol trials on 7 June 2021 in CENTRAL (Cochrane Controlled our primary method to analyse both direct and indirect
Register of Trials), CDSR (Cochrane Database of Systematic comparisons of ACEi/ARB (AA), BB, or both AABB, spi-
Reviews), PubMed and SCOPUS databases from database ronolactone (SPR) and statins, with placebo as the reference
inception until 7 June 2021. Our search was independently group. Standardised mean differences (SMD) were calcu-
executed by two reviewers (PJ and SS), and was based on lated from LVEF data and accompanying SDs extracted from
the Preferred Reporting Items for Systematic Reviews and each comparison.
712 P. Jeyaprakash et al.

Figure 1 PRISMA Flowchart. Literature search performed according to the Preferred Reporting Items for Systematic Re-
views and Meta-Analyses (PRISMA). After removing duplicates 4,007 records were screened by two independent reviewers
using Covidence software. Twelve studies were included in the network meta-analysis.

Our primary analysis used Bayesian inference with Monte agents from most effective to least effective at preserving
Carlo Markov chain-based sampling procedures. We used an LVEF. Meta-regression was performed to determine whether
uninformative prior determined by the gemtc package of R EAD, a previously reported determinant of ANT toxicity,
statistical software (The R Foundation for Statistical had a substantial influence towards cardioprotective effect.
Computing, Vienna, Austria) using the whole dataset. We Additionally, a frequentist method of network meta-analysis
used 5,000 burn-in iterations to check convergence, before was performed as a sensitivity analysis.
computing 100,000 actual simulation iterations to determine Risk of bias was assessed for all articles included in our
final probability estimates. A Gelman–Rubin plot and Gel- quantitative analysis using the Cochrane Risk of Bias
man diagnostics were used to assess the most optimal Monte assessment tool, which addressed issues with randomisation,
Carlo Markov chain-based model (Supplementary Figure 1). allocation concealment, blinding, attrition, and reporting bias
The node-split method was used to assess for inconsistency [18]. Heterogeneity between studies was assessed quantita-
between direct and direct comparisons. Our main analysis tively using the Cochrane I2 statistic [21]. Publication bias
compared all cardioprotective agents against placebo using was assessed using a funnel plot and Egger’s test. All
both direct and indirect comparisons. Standardised mean analysis was performed with R 4.1.0 statistical software [22],
differences (SMDs) in LVEF are presented with 95% credible using the “metafor", “dmetar” “gemtc”, “netmeta” and
intervals (CrIs). A subgroup analysis was performed studies “rjags” packages.
using AA as a reference group.
As a sensitivity analysis, we performed frequentist
network meta-analysis [20] to assess both direct and indirect Results
effects of each prophylactic drug class, where SMD with 95%
CI was reported. We also performed a subgroup frequentist A summary of the search is illustrated with a Preferred
analysis of studies where ANTs were used without trastu- Reporting Items for Systematic Reviews and Meta-Analyses
zumab to determine whether possible incremental car- flow diagram in Figure 1. Our structured search from four
diotoxicity from trastuzumab changed patient outcomes databases yielded 4,007 titles after removing duplicates. Af-
with prophylactic therapy. ter exclusion by title and abstract, 265 full-text articles were
Surface Under the Cumulative Ranking scores were assessed for eligibility. A total of 12 studies met all inclusion
calculated based on this analysis to rank cardioprotective criteria to be included in our network meta-analysis.
Preventing Anthracycline Cardiotoxicity: A Network Meta-Analysis 713

Table 1 Baseline study characteristics for included randomised controlled trials.

First author Year Country Follow- Anthracyclines Imaging Trastuzumab Trial arms (Drug1 n Mean ANT equivalent
up tool Target Dose) age, y dose mg/m2

Kaya [28] 2013 Turkey 6 months Dox or Epi TTE No Nebivolol 5 mg 27 51 257
Placebo 18 51 348
Liu [29] 2013 China 6 months Dox TTE No Carvedilol 5 mg 1 20 53 N/A
Candesartan 2.5 mg
Control 20 54 N/A
Elitok [37] 2014 Turkey 6 months Dox TTE No Carvedilol 12.5 mg 40 54 535
Placebo 40 53 523
Akpek [38] 2015 Turkey 6 months Dox or Epi TTE Yes Spironolactone 25 mg 43 50 482
Placebo 40 51 509
Pituskin [39] 2016 Canada 11 months Dox or Epi CMR Yes Bisoprolol 10 mg 31 53 592
Perindopril 8 mg 33 50 659
Placebo 30 51 643
Beheshti [30] 2016 Iran 1 week Dox TTE No Carvedilol 6.25 mg 30 42 240
Placebo 40 40 240
Gulati [9] 2016 Norway 4 months Epi CMR Yes Candesartan 32 mg1 28 50 297
Metoprolol 100 mg
Candesartan 32 mg 32 52 298
Metoprolol 50 mg 30 51 283
Placebo 30 51 283
Nabati [31] 2017 Iran 6 months Dox TTE No Carvedilol 6.25 mg 41 48 349
Placebo 40 47 360
Cochera [40] 2018 Romania 6 months Dox TTE No Nebivolol 5 mg 30 53 521
Control 30 52 519
Avila [41] 2018 Brazil 6 months Dox TTE No Carvedilol 50 mg 96 51 340
Placebo 96 53 338
Nabati [10] 2019 Iran 6 months Dox TTE Yes Rosuvastatin 20 mg 41 38 339
Placebo 40 39 338
Guglin [8] 2019 USA 12 months N/A TTE/ Yes Lisinopril 10 mg 65 51 N/A
MUGA Carvedilol 10 mg 59 52 N/A
Placebo 60 51 N/A

Anthracycline dose expressed in mg/m2 signifying dose/body surface area; ANT equivalent dose conversion as per European Society of Cardiology [19] -
(doxorubicin dose=idarubicin dose30.53=epirubicin dose30.7).
Abbreviations: N, number of participants in trial arm; ANT, anthracycline; Dox, doxorubicin; Epi, epirubicin; TTE, transthoracic echocardiography; N/A, Data
not available; CMR, cardiovascular magnetic resonance; MUGA, multigated acquisition scan.

The most common reasons for exclusion were paediatric/ Mean changes in LVEF pre and post ANT-based chemo-
animal studies (n=91), non-RCT data (n=30), and a lack of therapy from the trial arms of all analysed RCTs are reported
LVEF calculation both pre and post ANT exposure (n=48). in Table 2. We constructed a network plot (Figure 2) to
Baseline characteristics for the 12 included RCTs are illustrate direct and indirect comparisons of cardioprotective
shown in Table 1 and Supplementary Table 2. A total of 1,126 agents that were used from the included RCTs. We repre-
patients, all female, with a mean age of 50.5 years were sented the combination AA1BB (i.e., AABB in Figure 2) as a
included. A variety of cardioprotective agent comparisons separate node in our network plot to determine if there was
including BB alone (n=9), AA alone (n=3), Combination any additional synergistic effect on LVEF preservation.
BB1AA (n=2), SPR (n=1) and statin (n=1) were seen among A forest plot was generated to display the relative differ-
included trial arms. The weighted mean baseline LVEF was ence in mean LVEF for patients receiving cardioprotection,
62.6%, and the mean doxorubicin EAD was 412mg/m2, with the placebo arm being used as a reference (Figure 3).
dosed to body surface area. Ten trials used transthoracic None of the cardioprotective agents showed a statistically
echocardiography for LVEF assessment, while two used significant reduction in LVEF decline when compared to
cardiac magnetic resonance, and the mean follow-up LVEF placebo using the Bayesian model. The Gelman plot and
assessment occurred at 6 months. Gelman diagnostics used to assess the chosen Markov Chain
714 P. Jeyaprakash et al.

Table 2 Study outcomes of LVEF pre- and post-anthracycline-based chemotherapy.

First author Year Trial arms n LVEF pre6SD LVEF post6SD Mean difference6SD

Kaya [28] 2013 Nebivolol 5 mg 27 65.664.8 63.863.9 -1.865.8

Placebo 18 66.665.5 57.565.6 -9.1613.3
Liu [29] 2013 Carvedilol 5 mg1Candesartan 2.5 mg 20 74.468.6 57.562.6 -16.9619.4
Control 20 75.368.9 46.063.7 -29.3633.7
Elitok [37] 2014 Carvedilol 12.5 mg 40 65.064.5 63.364.8 -1.766.4
Placebo 40 66.066.1 64.165.1 -1.967.7
Akpek [38] 2015 Spironolactone 25 mg 43 67.066.1 65.767.4 -1.365.0
Placebo 40 67.766.3 53.666.8 -14625.1
Pituskin [42] 2016 Bisoprolol 10 mg 31 62.064.0 61.064.0 -1.065.0
Perindopril 8 mg 33 62.065.0 59.066.0 -364.0
Placebo 30 61.065.0 56.064.0 -5.065.0
Beheshti [30] 2016 Carvedilol 6.25 mg 30 61.363.2 61.163.4 -0.2260.9
Placebo 40 59.464.2 59.364.3 -0.2561.8
Gulati [9] 2016 Candesartan 32 mg1Metoprolol 100 mg 28 61.764.4 61.164.4 -0.663.9
Candesartan 32 mg 32 62.564.4 61.664.6 -0.963.9
Metoprolol 50 mg 30 63.364.3 60.864.2 -2.563.7
Placebo 30 63.164.4 60.364.7 -2.864.0
Nabati [10] 2017 Carvedilol 6.25 mg 41 58.764.7 57.467.5 -0.5565.6
Placebo 40 61.165.0 51.766.0 -9.565.9
Cochera [40] 2018 Nebivolol 5 mg 30 62.064.0 61.063.0 1.064.9
Control 30 61.062.0 60.063.0 1.063.5
Avila [41] 2018 Carvedilol 50 mg 96 64.864.7 63.963.8 -0.966.0
Placebo 96 65.263.6 63.965.2 -1.366.3
Nabati [10] 2019 Rosuvastatin 20 mg 38 55.164.8 53.566.7 -1.5166.6
Placebo 39 55.165.1 50.066.6 -5.265.7
Guglin [8] 2019 Lisinopril 10 mg 65 N/A N/A -460.8
Carvedilol 10 mg 59 N/A N/A -4.560.8
Placebo 60 N/A N/A -7.76 0.8

Abbreviations: LVEF, left ventricular ejection fraction; ANT, anthracycline; Mean difference, Post ANT mean–Pre ANT mean; SD, standard deviation; CI,
confidence interval (95%); n, number of participants in treatment arm; N/A, data not available.

Model are shown in Supplementary Figure 1. The node (Figure 3). The corresponding ranking of efficacy in the fre-
splitting analysis using the random effects model displayed quentist model is shown in Supplementary Table 4.
in Supplementary Figure 2 did not reveal inconsistency be- Our sensitivity analysis again demonstrated that the
tween direct and indirect comparisons between BB and AABB combination did not show a statistically substantial
AA1BB (p=0.70) or AA and AA1BB (p=0.58). Agents were protective effect. Statins and SPR were similarly not signifi-
then ranked according to relative efficacy, and results are cantly different compared to placebo under the frequentist
shown in Supplementary Table 3. model. A subgroup frequentist analysis only evaluating
Network meta-regression indicated no substantial rela- studies that used ANTs without trastuzumab also showed
tionship between EAD and cardioprotective agents (b=-0.08 no substantial difference with BBs compared to placebo
95% CrI -1.84, 1.73). Subgroup analysis results using AA as (Supplementary Figure 3).
the comparator arm are illustrated in the forest plot in Egger’s test performed to assess for publication bias did
Figure 4. In summary, there was no substantial LVEF pres- not indicate the presence of funnel plot asymmetry (-0.047
ervation in using BB (-0.23% 95% CrI -0.79, 0.36) or combi- [95% CI -3.19, 3.1]). A funnel plot for all direct comparisons
nation AA1BB (0.20 95% CrI -0.70, 1.1) when compared with between trials shown in Supplementary Figure 4 was also
AA. symmetrical.
However, our sensitivity analysis using the frequentist We used the Cochrane Risk of Bias assessment tool to
model revealed that AA and BB were able to significantly assess all included RCTs across five key domains, as shown
prevent LVEF decline when compared to placebo, with in Supplementary Table 5. The main potential sources of bias
SMDs of 1.32% (0.32,2.33) and 0.76% (0.12–1.40), respectively across all RCTs were a lack of description regarding
Preventing Anthracycline Cardiotoxicity: A Network Meta-Analysis 715

between different sonographers in many of the included

RCTs. Five of the included RCTs made no mention of
whether “intention to treat” analysis was performed and did
not report on the number of patients that crossed over if
there were clinical signs of cardiomyopathy.

Discussion
Our network meta-analysis provides an updated comparison
of cardioprotective agents in the setting of ANT-based
chemotherapy. Our primary analysis using Bayesian model
demonstrated that no substantial LVEF preservation was
offered by prophylactic use of any of the studied car-
dioprotective agents as compared to placebo. Therefore,
routine use of cardioprotective agent is not recommended
based on current literature. However, our sensitivity analysis
using the frequentist model showed that AA alone and BB
alone have substantial cardioprotective effects when
administered prophylactically. The discrepancy in the results
from the two statistical methods (i.e., Bayesian vs frequentist)
Figure 2 Network plot. Network plot showing direct
can, at least partially, explain the heterogeneity among the
and indirect drug class comparisons. Thicker lines
previous meta-analyses.
represent a higher number of direct comparisons.
Abbreviations: AA, angiotensin-converting enzyme There have been four network meta-analyses and one meta-
inhibitors/angiotensin receptor blockers; BB, beta- analysis previously performed examining cardioprotective
blockers; SPR, spironolactone, AABB, angiotensin re- agents. We have compared our study to these in Table 3,
ceptor blockers & beta-blockers. specifically highlighting the key differences we pre-specified
to provide new insights into the field of cardio-oncology.
The first of these published network meta-analyses assessing
randomisation process, and also failure to describe how
cardioprotection was in 2017, which examined cardiac out-
LVEF measurement was standardised across all included
comes rather than LVEF decline as a continuous variable [13].
patients. There was potential for inter-observer variability
Their results showed moderate evidence for both dexrazoxane

Figure 3 Overall effect of cardioprotective agents on LVEF relative to placebo. In Bayesian model (panel A), standardised
mean difference and 95% CrI are shown, whereas in frequentist model (panel B) standardised mean difference and 95% CI
are. The Bayesian model shows no significant difference in LVEF with the use of cardioprotective agents. However, in the
frequentist model, there is a statistically significant benefit with the use of ACE inhibitors/angiotensin receptor blockers or
beta-blockers. There was significant heterogeneity among studies, as evidenced by the frequentist model I2 statistic of 94%.
Abbreviations: ACE, angiotensin-converting enzyme; CI, confidence interval; CrI, credible interval; LVEF, left ventricular
ejection fraction.
716 P. Jeyaprakash et al.

Figure 4 Subgroup analysis: effect of cardioprotective agents on LVEF relative to ACEi/ARB using Bayesian Model. We
performed a subgroup analysis to determine whether there was any incremental difference between ACEi/ARB alone,
compared to combination ACEi/ARB1beta blockers, or against beta blockers alone. No difference was seen in either
comparison. Standardised mean difference and 95% credible interval are shown. Abbreviations: ACEi, angiotensin-
converting enzyme inhibitor; ARB, angiotensin receptor blocker; LVEF, left ventricular ejection fraction.

and AA, however, they indicated that more research is the latest publication [15]. Thus, a new meta-analysis is
required with larger RCTs to appropriately deal with the issue warranted. Second, the previous meta-analyses used only
of cardioprotection. While this provided important clinical one of two methods (Bayesian or frequentist). Performing
insights, the definitions of heart failure secondary to car- both Bayesian and frequentist meta-analysis allows us to
diotoxicity were inconsistent between the included studies, determine whether findings are statistically robust enough to
and as a result there was significant variability in the way inform clinical decision-making.
clinical heart failure outcomes were reported [23]. As a result, Both methods of analysis are valid and statistically
this review reported high rates of heterogeneity between trial accepted ways of performing meta-analysis. There are,
methodologies, and high risk of bias among many of their however, a few advantages to using a Bayesian approach
included studies. To account for this, we performed our meta- thus we chose it as the primary method. A Bayesian
analysis using the objective measure of SMD in LVEF rather approach does not assume that the testing parameter (in this
than using reports of heart failure. case, LVEF%) is fixed. Rather, it assumes that LVEF is
We have also designed and executed our network meta- random and has its own probability distribution, which can
analysis carefully to attempt to deal with key gaps and be modelled. This allows us to better model the variance of
areas of heterogeneity present in the recent meta-analyses the true effects (tau-squared) in the LVEF analysis. Bayesian
examining ANT cardioprotection published between 2017 analysis also uses the likelihood principle, which means that
and 2021 [13–16]. These trials included a less uniform patient any inferences are only based on available observed data,
population, as multiple cancer types were included, and rather than unobserved data in the frequentist analysis [24].
some non-randomised cohort studies were used for analysis. By showing that the statistical analysis method used is all
Paediatric studies were included in one of the meta-analyses that differentiates statistical significance, we are highlighting
[14], which resulted in more participant heterogeneity. One the fragility and uncertainty of published data in this space.
report diluted the strength of direct and indirect comparisons Interestingly combination AA/BB prophylaxis was not
by examining each drug individually rather than by drug shown to be effective even in our frequentist analysis, despite
class [15]. Meanwhile, another report condensed the effects AA and BB separately showing a signal towards benefit in
by combining all the cardioprotective agents together [16]. the frequentist analysis. One of the explanations for this may
These prior reviews importantly included trials where pa- be the low overall numbers that received AABB combination
tients had in some cases received prior ANT exposure. Un- across included trials (n=48). Results previously published
like this review, neither of the prior reviews performed meta- by our group have shown that ANT cardiotoxicity is much
regression to determine whether ANT dose was a less than previously reported using modern ANT regimens,
confounder for their results. with a weighted mean reduction in LVEF of only 5.4% in
There are three key reasons that motivated another meta- placebo arms [25]. Therefore, the number of participants
analysis. First, the previous meta-analyses were discordant, included across the studies in this meta-analysis may not be
resulting in confusion at the highest level of evidence. large enough to detect a statistically significant benefit in
Several years have passed since the last search (June 2019) of combination AABB therapy. Another possibility is that there
Preventing Anthracycline Cardiotoxicity: A Network Meta-Analysis 717

Table 3 Summary of previous network meta-analysis involving cardioprotective agents.

Items Our Study Abdel-Qadir [13] Vaduganathan [16] Li [14] Alizadehasl [15]

Publication 2024 2017 2019 2020 2021

year
Number of 12 16 17 22 12
articles
Primary Change in LVEF Cardiac Outcome Change in LVEF Change in LVEF Change in LVEF
outcome (Mean Difference) (Odds Ratios) (Mean Difference) (Mean Difference) (Mean Difference)
Search date June 2021 January 2015 March 2018 December 2018 June 2019
completion
Study type Extensively peer RCTs, not all RCTs, not all Combination of study RCTs, not all extensively
reviewed RCTs only extensively peer extensively peer types including RCTs, peer reviewed e.g.,
reviewed e.g., reviewed e.g., included retrospective cohorts included grey literature
included grey grey literature and prospective
literature cohorts
Inclusion Very strict criteria that Studies not Studies not necessarily Studies not Studies report primary
criteria studies only included necessarily involving primary necessarily involving prevention of CTRCD
anthracycline naïve involving primary prevention included primary prevention exclusively
patients prevention included e.g., Cardinale et. al. included e.g., Dessi et.
e.g., Dessi. al.
Paediatric studies
excluded.
Exclusion Paediatric patients Paediatric patients Paediatric patients Paediatric patients Paediatric patients
excluded. Subgroup excluded excluded included e.g., Gupta excluded
RCTs e.g., cardinal 2006 2017
where only troponin 1ve
patients were
randomised were
excluded.
Malignancy Breast cancer only Mix of solid organ/ Breast cancer, Breast cancer, Breast cancer or
type haematological Hodgkins, non- Hodgkins, non- Lymphoma
malignancies (66% Hodgkins, leukemia, Hodgkins, leukemia,
breast cancer 19% endometrial cancer endometrial cancer
haematological)
Outcome LVEF change from Study dependent LVEF change from LVEF change from LVEF change from
assessed baseline definitions of baseline baseline baseline
outcome e.g., heart
failure or LVEF
decline .10%.
Prior No Yes in some cases. Yes in some cases Yes in some cases Yes in some cases
anthracycline
exposure
Statistical Both Bayesian analysis Clinical Outcomes. Frequentist meta- Frequentist Network Frequentist Network
analysis and frequentist analysis Bayesian Network analysis Meta-analysis in Stata Meta-analysis in R
method performed in R Software Meta-analysis in using SMD Software
OpenBUGS
software
Medication Class Class Multiple classes Class and drug Drug
drug or Class combined for overall
effect neurohormonal effect
718 P. Jeyaprakash et al.

Table 3. (continued).

Items Our Study Abdel-Qadir [13] Vaduganathan [16] Li [14] Alizadehasl [15]

Metaregression Yes (Continuous Yes (categorical No No No

against variable) variable)
anthracycline
dose
Study Age Studies included between Studies ranging Studies included Studies between 2006 Studies included between
2010 and 2019 from 1992 to 2013 between 2006 and 2019 and 2018 2010 and 2018
Conclusions No clear superiority Dexrazoxane and Small but statistically Most agents including Spironolactone was the
shown with any one AA are most likely significant benefit when ACE inhibitors, only drug to show
cardioprotective agent in to be effective and neurohormonal therapy statins, BB, statistically significant
Bayesian analysis. preventing effects were collectively spironolactone benefit, however this was
Frequentist analysis cardiotoxicity and examined, when showed statistically based on a single small
showed significant adverse cardiac compared to control. significant study. Metoprolol was
benefit for BB and AA. outcomes. cardioprotective not recommended for
benefits. cardioprotection.

Abbreviations: LVEF, left ventricular ejection fraction; RCT, randomized controlled trial; CTRCD, cancer therapeutics-related cardiac dysfunction; SMD,
standardised mean difference; AA, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker; ACE, angiotensin-converting enzyme; BB, beta-
blocker.

may be an unknown counterproductive mechanism when demonstrated that isolated AA and isolated BB were car-
both AA and BB are concurrently prescribed, however, this dioprotective rather than combination therapy as suggested
seems unlikely given the large body of evidence for combi- by the guidelines [12].
nation AABB therapy in the setting of heart failure with
systolic dysfunction.
We identified significant variability in the choices of BB and Limitations
AA between our included studies, as four different BB (car-
vedilol, metoprolol, nebivolol, or bisoprolol), and three AA One of the limitations of the assessment of ANT-induced
agents (candesartan, perindopril, lisinopril) were used. The cardiotoxicity in our included studies is that changes in
biochemical properties of carvedilol in particular have been LVEF measured using traditional echocardiography occur
examined in vitro, and there is evidence to indicate additional late. Paediatric data has shown large variations in the onset
antioxidant properties when compared to other BBs [26]. This of cardiotoxicity, from 1 week up to 20 years after ANT
may have had a clinically significant effect on our analysis, as administration, which are independent of established risk
the proposed mechanism of ANT cardiotoxicity is the pro- factors of cumulative dose and pre-existing cardiac disease
duction of reactive oxygen species. There were also differences [33]. The majority of RCTs we included only reported follow-
in BB dose, with carvedilol doses ranging from the lowest up periods of 6 months, which may have been insufficient to
6.25mg dose up to the highest 50mg dose. Heart failure detect late cardiotoxicity. Studies using more sensitive mea-
guidelines indicate that in order to maximise the benefit of BB sures of cardiac function such as myocardial strain have
and AA, up-titration to the maximal tolerated dose is required shown that adverse changes can develop months before
[27]. Some of the included studies [28–31] may not have changes in LVEF occur [34]. There is a proposed large trial
adequately up-titrated drugs to achieve maximal protective evaluating the role of BBs in preventing cardiotoxicity using
effect prior to ANT administration, possibly underestimating the outcome of changes myocardial strain instead of LVEF,
their role in preventing LVEF decline. and this may provide important insights [35].
The guidelines currently encourage the use of the Heart Despite our efforts to study a homogeneous cohort by
Failure Association and International Cardio-Oncology So- learning from previous reviews, there was still marked het-
ciety risk assessment tool to stratify cardiotoxicity risk erogeneity identified between included trial arms. There was
[12,32]. Those in the high and very high-risk groups are then a wide variation in LVEF decline, particularly among the
encouraged to commence prophylactic cardioprotection with placebo arms, which implied that some trial populations
AA and BB [12]. Although our study does not recommend may be more susceptible to ANT cardiotoxicity than others.
blanket cardioprotection prescription prior to ANT therapy, In addition, the protocols for chemotherapy, cardioprotective
it does potentially support the latest guidelines by suggest- drugs and LVEF measurement varied significantly among
ing a certain population may benefit [12]. However, it should included studies. To mitigate this, we only included RCTs in
be noted that our study in the frequentist model our analysis to minimise inherent bias.
Preventing Anthracycline Cardiotoxicity: A Network Meta-Analysis 719

There was also significant variability in the individual BB Appendices

and ACEi drugs used between trials. As a result, our analysis
investigated the class effect of these agents. More studies are Supplementary data associated with this article can be
required to determine whether individual drugs within the found, in the online version, at https://doi.org/10.1016/j.
same class would result in different outcomes. The variability hlc.2023.11.004.
in drug dosing and protocols regarding drug up-titration also
made it difficult to infer any association between drug dose
and degree of cardioprotection. A meta-regression analysis of
References
drug dosing could not be performed due to this variability. [1] Weiss RB. The anthracyclines: will we ever find a better doxorubicin?
Semin Oncol. 1992;19(6):670–86.
Studies studying the effect of cardioprotection agents with a
[2] Hortobágyi GN. Anthracyclines in the treatment of cancer. An overview.
clear titration protocol would be useful to determine if car- Drugs. 1997;54(Suppl 4):1–7.
dioprotection is dose-dependent. For the fair comparisons [3] Swain SM, Whaley FS, Ewer MS. Congestive heart failure in patients
treated with doxorubicin: a retrospective analysis of three trials. Cancer.
among the studies within the model, SMD was used in our
2003;97(11):2869–79.
meta-analysis. Of note, the point estimates from the absolute [4] Armenian SH, Lacchetti C, Barac A, Carver J, Constine LS, Denduluri N,
SMD values does not indicate the mean difference between the et al. Prevention and monitoring of cardiac dysfunction in survivors of
adult cancers: American Society of Clinical Oncology clinical practice
two drugs (i.e., not the difference in actual EF). Rather, it in-
guideline. J Clin Oncol. 2017;35(8):893–911.
dicates the order of the effect sizes when compared protective [5] Lotrionte M, Biondi-Zoccai G, Abbate A, Lanzetta G, D’Ascenzo F,
effect of one drug to another. Many of the included RCTs did Malavasi V, et al. Review and meta-analysis of incidence and clinical
predictors of anthracycline cardiotoxicity. Am J Cardiol.
not report extensive baseline comorbidities of their trial pop-
2013;112(12):1980–4.
ulation. In particular, the distribution of risk factors for [6] Jang S, Ruddy KJ, Loprinzi CL, Armenian S, Lenihan DJ, Herrmann J.
developing cardiotoxicity was not available for us to perform Systematic Review and Meta-analysis of Definitions and Reported In-
cidences of Anthracycline Cardiotoxicity. American Society of Clinical
subgroup analysis. While we were able to correct for ANT
Oncology; 2019. https://ascopubs.org/doi/10.1200/JCO.2019.37.15_
dose and age, comorbidities such as diabetes, hypertension, suppl.e23081. [accessed 12.10.23].
and obesity (which have been demonstrated to increase [7] Boekhout AH, Gietema JA, Milojkovic Kerklaan BM, van Werkhoven ED,
Altena R, Honkoop A, et al. Angiotensin II-receptor inhibition with
cardiomyopathy risk), could not be evaluated [5,36]. Larger
candesartan to prevent trastuzumab-related cardiotoxic effects in patients
trials, or an individual patient-level meta-analysis with further with early breast cancer: a randomized clinical trial. JAMA Oncol.
data would be required to determine whether certain 2016;2(8):1030–7.
[8] Guglin M, Krischer J, Tamura R, Fink A, Bello-Matricaria L, McCaskill-
higher-risk populations would benefit more from
Stevens W, et al. Randomized Trial of Lisinopril Versus Carvedilol to
cardioprotective agents. Prevent Trastuzumab Cardiotoxicity in Patients With Breast Cancer. J Am
Coll Cardiol. 2019;73(22):2859–68.
[9] Gulati G, Heck SL, Ree AH, Hoffmann P, Schulz-Menger J,
Fagerland MW, et al. Prevention of cardiac dysfunction during adjuvant
Conclusion breast cancer therapy (PRADA): a 2 3 2 factorial, randomized, placebo-
controlled, double-blind clinical trial of candesartan and metoprolol.
There is no sufficient evidence to support prophylactic pre- Eur Heart J. 2016;37(21):1671–80.
[10] Nabati M, Janbabai G, Esmailian J, Yazdani J. Effect of rosuvastatin in
scription of cardioprotective agents to prevent EF reduction preventing chemotherapy-induced cardiotoxicity in women with breast
for patients with breast cancer and receiving potentially cancer: a randomized, single-blind, placebo-controlled trial. J Cardiovasc
cardiotoxic chemotherapy. However, frequentist analysis Pharmacol Ther. 2019;24(3):233–41. D.
[11] Salehi R, Zamani B, Esfehani A, Ghafari S, Abasnezhad M, Goldust M.
suggested that AA or BBs provide cardioprotection. Thus, Protective effect of carvedilol in cardiomyopathy caused by anthracy-
the different results from the two methods can, at least clines in patients suffering from breast cancer and lymphoma. Am Heart
partially, explain the heterogeneity among the previous Hosp J. 2011;9(2):95–8.
[12] Lyon AR, López-Fernández T, Couch LS, Asteggiano R, Aznar MC,
meta-analyses. Besides, there are concerns with heterogene- Bergler-Klein J, et al. ESC Scientific Document Group. 2022 ESC Guide-
ity in trial design, lack of conclusive outcome data, and lines on cardio-oncology developed in collaboration with the European
variability in drug dosage that limit the interpretation of our Hematology Association (EHA), the European Society for Therapeutic
Radiology and Oncology (ESTRO) and the International Cardio-
findings. Ultimately, larger scale studies with standardised Oncology Society (IC-OS). Eur Heart J. 2022;43(41):4229–361.
protocols and adequate drug up-titration are required to [13] Abdel-Qadir H, Ong G, Fazelzad R, Amir E, Lee DS, Thavendiranathan P,
provide further insight into whether cardioprotection is et al. Interventions for preventing cardiomyopathy due to anthracyclines:
a Bayesian network meta-analysis. Ann Oncol. 2017;28(3):628–33.
effective, and which agent should be used. [14] Li X, Li Y, Zhang T, Xiong X, Liu N, Pang B, et al. Role of cardioprotective
agents on chemotherapy-induced heart failure: a systematic review and
network meta-analysis of randomized controlled trials. Pharmacol Res.
Funding Sources 2020;151:104577.
[15] Alizadehasl A, Ghadimi N, Kaveh S, Maleki M, Ghavamzadeh A, Noohi F,
This research did not receive any specific grant from funding et al. Prevention of anthracycline-induced cardiotoxicity: a systematic re-
view and network meta-analysis. Int J Clin Pharm. 2021;43(1):25–34.
agencies in the public, commercial, or not-for-profit sectors.
[16] Vaduganathan M, Hirji SA, Qamar A, Bajaj N, Gupta A, Zaha VG, et al.
Efficacy of neurohormonal therapies in preventing cardiotoxicity in pa-
tients with cancer undergoing chemotherapy. JACC CardioOncology.
2019;1(1):54–65.
Competing Interest Statement [17] Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L,
et al. Recommendations for cardiac chamber quantification by echocar-
There are competing interests to disclose. diography in adults: an update from the American Society of
720 P. Jeyaprakash et al.

Echocardiography and the European Association of Cardiovascular Im- [31] Nabati M, Janbabai G, Baghyari S, Esmaili K, Yazdani J. Cardioprotective
aging. Eur Heart J Cardiovasc Imaging. 2015;16(3):233–70. effects of carvedilol in inhibiting doxorubicin-induced cardiotoxicity.
[18] Collaboration C. Rev Manag (RevMan): Oxford 2015;5(3. 5). J Cardiovasc Pharmacol. 2017;69(5):279–85.
[19] Plana JC, Galderisi M, Barac A, Ewer MS, Ky B, Scherrer-Crosbie M, et al. [32] Lyon AR, Dent S, Stanway S, Earl H, Brezden-Masley C, Cohen-Solal A,
Expert consensus for multimodality imaging evaluation of adult patients et al. Baseline cardiovascular risk assessment in cancer patients scheduled
during and after cancer therapy: a report from the American Society of to receive cardiotoxic cancer therapies: a position statement and new risk
Echocardiography and the European Association of Cardiovascular Im- assessment tools from the Cardio-Oncology Study Group of the Heart
aging. Eur Heart J Cardiovasc Imaging. 2014;15(10):1063–93. Failure Association of the European Society of Cardiology in collabora-
[20] Shim SR, Kim SJ, Lee J, Rücker G. Network meta-analysis: application tion with the International Cardio-Oncology Society. Eur J Heart Fail.
and practice using R software. Epidemiol Health. 2019;41:e2019013. 2020;22(11):1945–60.
[21] Borenstein M, Higgins JPT, Hedges LV, Rothstein HR. Basics of meta- [33] Liesse K, Harris J, Chan M, Schmidt ML, Chiu B. Dexrazoxane signifi-
analysis: I2 is not an absolute measure of heterogeneity. Res Synth cantly reduces anthracycline-induced cardiotoxicity in pediatric solid
Methods. 2017;8(1):5–18. tumor patients: a systematic review. J Pediatr Hematol Oncol.
[22] R Core Team. R: A Language and Environment for Statistical Computing. 2018;40(6):417–25.
Vienna, Austria: R Foundation for Statistical Computing; 2018. https:// [34] Thavendiranathan P, Poulin F, Lim KD, Plana JC, Woo A, Marwick TH.
www.scirp.org/reference/ReferencesPapers?ReferenceID=2342186. Use of myocardial strain imaging by echocardiography for the early
[accessed 12.10.23]. detection of cardiotoxicity in patients during and after cancer
[23] Gujral DM, Lloyd G, Bhattacharyya S. Effect of prophylactic betablocker chemotherapy: a systematic review. J Am Coll Cardiol. 2014;63(25A):
or ACE inhibitor on cardiac dysfunction & heart failure during anthra- 2751–68.
cycline chemotherapy 1/- trastuzumab. Breast (Edinb Scotl). 2018;37:64– [35] Negishi T, Thavendiranathan P, Negishi K, Marwick TH. SUCCOUR
71. Investigators. Rationale and design of the strain surveillance of chemo-
[24] Jack Lee J, Chu CT. Bayesian clinical trials in action. Stat Med. therapy for improving cardiovascular outcomes: the SUCCOUR trial.
2012;31(25):2955–72. J Am Coll Cardiol Imaging. 2018;11(8):1098–105.
[25] Jeyaprakash P, Sangha S, Ellenberger K, Sivapathan S, Pathan F, [36] Guenancia C, Lefebvre A, Cardinale D, Yu AF, Ladoire S, Ghiringhelli F,
Negishi K. Cardiotoxic effect of modern anthracycline dosing on left et al. Obesity as a risk factor for anthracyclines and trastuzumab car-
ventricular ejection fraction: A systematic review and meta-analysis of diotoxicity in breast cancer: a systematic review and meta-analysis. J Clin
placebo arms from randomized controlled trials. J Am Heart Assoc. Oncol. 2016;34(26):3157–65.
2021;10(6):e018802. [37] Elitok A, Oz F, Cizgici AY, Kilic L, Ciftci R, Sen F, et al. Effect of carvedilol
[26] Spallarossa P, Garibaldi S, Altieri P, Fabbi P, Manca V, Nasti S, et al. on silent anthracycline-induced cardiotoxicity assessed by strain imaging:
Carvedilol prevents doxorubicin-induced free radical release and a prospective randomized controlled study with six-month follow-up.
apoptosis in cardiomyocytes in vitro. J Mol Cell Cardiol. 2004;37(4):837– Cardiol J. 2014;21(5):509–15.
46. [38] Akpek M, Ozdogru I, Sahin O, Inanc M, Dogan A, Yazici C, et al. Pro-
[27] Hunt SA, Baker DW, Chin MH, Cinquegrani MP, Feldman AM, tective effects of spironolactone against anthracycline-induced cardio-
Francis GS, et al. ACC/AHA guidelines for the evaluation and man- myopathy. Eur J Heart Fail. 2015;17(1):81–9.
agement of chronic heart failure in the adult: executive summary. J Am [39] Pituskin E, Haykowsky M, McNeely M, Mackey J, Chua N, Paterson I.
Coll Cardiol. 2001;38(7):2101–13. Rationale and design of the multidisciplinary team IntervenTion in
[28] Kaya MG, Ozkan M, Gunebakmaz O, Akkaya H, Kaya EG, Akpek M, cArdio-oNcology study (TITAN). BMC Cancer. 2016;16(1):733.
et al. Protective effects of nebivolol against anthracycline-induced car- [40] Cochera F, Dinca D, Bordejevic DA, Citu IM, Mavrea AM, Andor M, et al.
diomyopathy: a randomized control study. Int J Cardiol. Nebivolol effect on doxorubicin-induced cardiotoxicity in breast cancer.
2013;167(5):2306–10. Cancer Manag Res. 2018;10:2071–81.
[29] Liu L, Liu ZZ, Liu YY, Zheng ZD, Liang XF, Han YL, et al. [Preventive [41] Avila MS, Ayub-Ferreira SM, de Barros Wanderley MR Jr, das Dores
effect of low-dose carvedilol combined with candesartan on the car- Cruz F, Gonçalves Brandão SM, Rigaud VOC, et al. Carvedilol for pre-
diotoxicity of anthracycline drugs in the adjuvant chemotherapy of breast vention of chemotherapy-related cardiotoxicity: the CECCY trial. J Am
cancer]. Zhonghua Zhong Liu Za Zhi [Chin J Oncol]. 2013;35(12):936–40. Coll Cardiol. 2018;71(20):2281–90.
[30] Tashakori Beheshti A, Mostafavi Toroghi H, Hosseini G, Zarifian A, [42] Pituskin E, Mackey JR, Koshman S, Jassal D, Pitz M, Haykowsky MJ, et al.
Homaei Shandiz F, Fazlinezhad A. Carvedilol administration can prevent Multidisciplinary approach to novel therapies in cardio-oncology research
doxorubicin-induced cardiotoxicity: a double-blind randomized trial. (MANTICORE 101-breast): a randomized trial for the prevention of
Cardiology. 2016;134(1):47–53. trastuzumab-associated cardiotoxicity. J Clin Oncol. 2017;35(8):870–7.