Article Text

Myocardial disease
How to monitor cardiac toxicity of chemotherapy: time is muscle!
  1. D Kerkhove1,
  2. C Fontaine2,
  3. S Droogmans1,
  4. J De Greve2,
  5. K Tanaka3,
  6. Nico Van De Veire1,4,
  7. Guy Van Camp1
  1. 1Centrum voor Hart, en Vaatziekten, Universitair Ziekenhuis Brussel Laarbeeklaan, Brussels, Belgium
  2. 2Borstkliniek, Oncologisch centrum, Universitair Ziekenhuis Brussel, VUB, Jette, Brussels, Belgium
  3. 3Department of Radiology, Universitair Ziekenhuis Brussel—VUB, Jette, Brussels, Belgium
  4. 4Algemeen ziekenhuis Maria Middelares, Ghent, Belgium
  1. Correspondence to Dr D Kerkhove, Centrum voor Hart- en Vaatziekten, Universitair Ziekenhuis Brussel Laarbeeklaan 101, Jette 1090, Belgium; Dirk.kerkhove{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Cancer and cardiovascular (CV) disease are the leading causes of death in the western world. The increasing age of the western population and the improved global cancer survival, due to ongoing advances in anticancer treatments, point to further increases in the CV burden in cancer survivors as a consequence of exposure to cardiotoxic drugs. Since it is not desirable to improve cancer survival at the price of increased CV mortality and morbidity, more and more cardio-oncological papers are seeking methods to enhance the early diagnosis of cardiotoxicity.

Before being able to determine which tests are best at identifying cardiotoxicity in a quick, accurate and prognostically relevant way, we have to understand how cardiotoxicity has been defined.

Lack of consensus in defining cardiotoxicity

Histological proof of cardiotoxicity remains the diagnostic gold standard, but its correlation with functional cardiac status is unclear, and additional practical limitations make it unsuitable for detecting cardiotoxicity.

Historically, left ventricular ejection fraction (LVEF) was chosen as the parameter for monitoring/identifying cardiotoxicity in clinical practice with proven impact on prognosis. LVEF expresses global LV systolic function, it correlates well with functional status for the majority of the population, and several well studied methods can measure LVEF with acceptable to high accuracy. However, its limitations have been well established over the last 10 years, ranging from methodological through mathematical to statistical ones.

Cardiotoxicity was first defined by Schwartz et al1 in an anthracycline treated population, using multiple gated acquisition (MUGA) scanning to determine LVEF during and after anthracycline treatment. The diagnosis of cardiotoxicity was fulfilled in the case of an absolute 10% drop in LVEF or a decrease to below 50% in patients with baseline LVEF >50%, or an absolute 10% drop in LVEF or a drop below 30% in patients with baseline LVEF <50%. Several years later, the Cardiac Review and Evaluation Committee (CREC) for trastuzumab trials, a chemotherapeutic monoclonal antibody directed against human epidermal growth factor 2 receptor (HER-2) expressing breast cancer, has formulated a new standard for cardiotoxicity, containing the following criteria: (1) a decline in LVEF, either global or more severe in the septum; (2) symptomatology of congestive heart failure (CHF); (3) associated signs of CHF; (4) a symptomatic decline in LVEF of at least 5% to a final value of <55%; or (5) an asymptomatic decline in LVEF of at least 10% to a final value of <55%.w1 Many papers use a clinical definition: (1) an asymptomatic drop in LVEF ≥10% below 55%; (2) an asymptomatic but lesser drop in LVEF <55%; and (3) a symptomatic but lesser drop in LVEF.w2

In conclusion, no consensus currently exists about how cardiotoxicity should be exactly defined. As a consequence, comparison of the different methods for serial LVEF monitoring to detect cardiotoxicity, evaluated within and across studies, is difficult.

Limitations of serial LVEF as a cardiotoxicity defining parameter

Serial LVEF monitoring has been accepted but never validated for identifying cardiotoxicity. Its most important limitation is the fact that the LVEF drop comes at the end of the pathophysiological cascade.w3 At that point, cardiac damage is often irreversible, almost excluding a window for the prevention of cardiotoxicity. Furthermore, it has been shown that, even if LVEF decreases during treatment, it remains within normal limits when averaged for a whole group in large studies.w4 Accordingly, it should be emphasised that many patients develop histological evidence of myocardial changes without developing resting LV systolic dysfunction (LVSD).2

Accuracy and reproducibility of echocardiographic 2D Simpson's method (biplane EF (BPEF)) are limited by a number of factors including image quality (especially endocardial border delineation) and alignment difficulties in apical views, or non-universally applicable geometrical assumptions in the 2D LV model, stressing its known operator dependency. Furthermore, small or large LV volumes with comparable contractile state influence LVEF calculation in the opposite direction due to mathematics only. Finally, LVEF is also susceptible to changes in loading. The summation of all these factors leads to a minimal detectable change in BPEF of up to 11%.3

The quest to improve accuracy and reproducibility of LVEF measurement

3D echocardiography and triplane Simpson's method

3D echocardiography (3DE) consistently improves accuracy and reproducibility of LVEF measurement compared to biplane and triplane methods, by abolishing the need for geometrical assumptions. It has been shown to correlate more closely with MRI derived LVEF than 2D echocardiography.4 ,5 In a small (N=56) breast cancer population receiving chemotherapy but with cardiotoxicity excluded on the basis of stable global longitudinal strain (GLS), Thavendiranathan calculated the standard error of measurement or temporal variability for BPEF, triplane or 3DE as being, respectively, 4.9% (95% CI 4.5% to 5.4%), 5.9% (95% CI 5.4% to 6.5%), and 2.8% (95% CI 2.5% to 3.1%).6 This error contributed to minimal detectable changes in LVEF of, respectively, 13.0%, 16.2%, and 6.0%. This illustrates that adding one dimension in the triplane LVEF calculation does not improve accuracy. Triplane acquisition is more prone to off-axis views, endocardial border delineation is inferior to 2D, and the lower acquired frame rate increases the error in identifying the real end-systolic and end-diastolic volume frames.6

‘Yes’ for (semi-)automated endocardial border detection, ‘No’ for contrast use

The operator dependency of LVEF calculation can be limited by applying highly reproducible software controlled (semi-)automated endocardial border tracking. It has proven to increase reproducibility of LVEF calculation compared to manual contouring, both with 2Dw5 and 3D.6 w6

Although the use of contrast for LV opacification improves endocardial border delineation, according to the literature it has not consistently improved accuracy and reproducibility, either for BPEF,w7–w10 or for triplane Simpson's method or 3DE.6 The main reason is that the machine settings for contrast use exclude the ability of automated endocardial border detection, thereby confirming the importance of this feature.

Stress echocardiography

The role of stress echocardiography in identifying cardiotoxicity is unclear because: (1) data are scarce and mainly cross-sectional; (2) tested populations have been small; and (3) results have been conflicting. Two studies, both concerning a paediatric population, showed an abnormal response of systolic LV function to exercise,w11 w12 but others could not confirm this finding.w13 w14

Cardiac magnetic resonance

Because of its excellent reproducibility and accuracy,7 cardiac magnetic resonance (CMR) derived volumes and LVEF are considered the imaging gold standard for any myocardial disorder, chemotherapy related LVSD included.8 ,9 Its high cost and limited availability precludes its first line use for serial LVEF monitoring in clinical routine, but MRI should be considered as the first alternative to echocardiography in the case of poor echogenicity.

MUGA scanning

As CMR was still in development when the first algorithms for detecting cardiotoxicity were introduced into oncological practice in the 1980s, MUGA was considered the gold standard for serial LVEF calculation at that time, because of its high reproducibility/low variability1 compared to BPEF. Strashun's study evaluating the variability in 101 patients with two different acquisition protocols found the intra-observer variability and test–retest variability to be 2.9% and 3.4%, respectively.w15 In very recent papers, it withstands comparison with 3DE and shows comparable correlation with CMR LVEF, even in serial monitoring (r=0.88, r=0.97, r=0.87 for MUGA compared to r=0.91, r=0.97, r=0.90 for 3DE at baseline, 6, and 12 months, respectively).5

There are many clinical drawbacks with MUGA scanning: low reproducibility in patients with arrhythmia because of interference with the averaging process, high cost compared to echocardiography, and evaluation of LV volumes and LVEF only in routine acquisition. More importantly, MUGA increases the cumulative radiation dose significantly, especially when used for serial monitoring (7.8 mSv/scan).w16

Monitoring cardiac toxicity of chemotherapy: key points 1

  • The generally improved cancer survival points to an increased cardiovascular burden in cancer survivors.

  • There is no actual consensus among cardio-oncologists on how cardiotoxicity should be exactly defined.

  • Even though minimal detectable change in biplane LVEF is >10%, actual clinical practice in detecting cardiotoxicity is almost exclusively based on biplane LVEF monitoring.

  • Since an LVEF decrease is a late and insensible marker for cardiotoxicity, monitoring LVEF only excludes almost completely the time window for cardio-oncological prevention.

  • MRI is the gold standard for serial LVEF monitoring, but its high cost and limited availability preclude its use as the first line method.

  • Echocardiography remains the first line method because it offers a complete and integrated evaluation of cardiac function: systolic and diastolic ventricular function as well as valvular and pericardial morphology and function.

  • In the case of good echogenicity, 3D LVEF approaches the best MRI performance.

  • In the case of poor echogenicity, echocardiography is best replaced by cardiac MR instead of multiple gated acquisition (MUGA). It avoids the additional irradiation from MUGA and also provides information on valvular and pericardial morphology and function.

Beyond LVEF: techniques trying to enhance the identification of cardiotoxicity

The ultimate goal of cardio-oncology is to prevent cardiotoxicity so that the oncologic prognosis can be further improved without worsening the cardiac prognosis. The weaknesses of LVEF as a qualifying parameter for cardiotoxicity are a clear obstacle to reaching this goal. Understandably, the literature has focused on newer methods that try to identify myocardial dysfunction in an earlier, preclinical stage (ie, before LVEF declines) without losing prognostic impact.

Diastolic dysfunction in the context of cardiotoxicity

Diastolic dysfunction occurs frequently in patients receiving chemotherapy, but its incidence is also high in the ‘normal healthy population’, especially with increasing age.w17 Consequently, most of the studies evaluating its ability to predict chemotherapy related LVSD were negative,10 apart from one.w18

Deformation imaging

The potentially high diagnostic power of deformation imaging was first illustrated by Ganamé et al in a paediatric population receiving high doses of anthracyclines for haematologic malignancies; the first dose reduced peak strain and strain rate and predicted later LVSD.w19 Many papers have been published meanwhile, testing various types of global and regional deformation parameters, with rather short follow-up times in mostly small populations treated mainly for breast cancer with various therapeutic regimens. Not surprisingly, data on the predictive value of most deformation parameters are conflicting. Furthermore, it is still unclear which deformation parameter is best for identifying cardiotoxicity at an early stage. The main reason is that we still do not know if chemotherapy affects myocardial contractility in the same way as it does myocardial ischaemia, aortic stenosis or arterial hypertension—that is, by first reducing longitudinal function (of the subendocardial fibres) before radial/circumferential function (of the mid myocardial and epicardial fibres). Some studies identified simultaneously reduced longitudinal, radial and circumferential strain, suggesting that chemotherapy affects the whole myocardial thickness homogeneously.w20–w22 Other studies, in contrast, found that changes in radial and circumferential deformation appeared later than longitudinal changes,11 w23 or were non-significant,12 confirming that longitudinal LV mechanics are most vulnerable and sensitive in the development of myocardial disease.

An alternative explanation would be the clinically experienced difficulties in obtaining confident radial and circumferential deformation curves, affecting their reproducibility. This statement is supported by the JUSTICE study,13 which determined normal range 2D strain values and their inter/intra-observer and inter-vendor differences (GE Vivid 7 or E9, Philips iE33 and Toshiba Artida or Aplio) in a normal population (N=817). It confirmed that variation in global radial strain (GRS) values, expressed as 1 SD, is much higher (8<SD <12%) compared to global longitudinal strain (GLS) and global circumferential strain (GCS) variation (2<SD<4%).

Global longitudinal strain with 2D strain (speckle tracking)

Mignot et al have proven that GLS is the best prognosticator for global cardiac mortality in the case of reduced systolic function (N=147) in the general population; its sensitivity and specificity were, respectively, 73% and 83% (AUC=0.83) compared to 73% and 58% for LVEF (AUC=0.72).14 Particularly in those subjects with normal or near normal LVEF, GLS was found to add prognostic information, even if 3D LVEF predicted event-free survival better than GLS (χ2=25.4; p<0.001 for 3D LVEF vs χ2=21.7; p<0.001 for GLS vs χ2=19.5; p<0.001 for 2D LVEF).w24 In the specific context of cardiotoxic treatment, Negishi et al confirmed that a reduced GLS of 11% was the strongest predictor (sensitivity 65%, specificity 94%, AUC=0.84) and also an independent predictor of subsequent cardiotoxicity (expressed as a 10% decline in LVEF) in 81 breast cancer patients, even after adjustment for clinical risk factors for cardiotoxicity such as hypertension, smoking, diabetes, and advanced age.12

Moreover, GLS's diagnostic power for detecting subtle changes in myocardial systolic function during and after cardiotoxic treatment has been sharply illustrated in several but small cross-sectional studies. Ho et al confirmed that smoking enhances cardiotoxicity10; while GLS was normal and identical for smokers and non-smokers in the non-treated control group (respectively −20±2% vs −20±1%; p=1.0), it dropped significantly in the smokers, compared to the non-smokers, after chemotherapy (respectively −18±2% vs 16.0±2%; p=0.01). In parallel with Negishi's findings, a significant 10% reduction in GLS (from baseline) after 3 months predicted a future reduction in LVEF (79% sensitivity, 82% specificity). Sawaya et al confirmed those findings in a similar protocol,w25 leading to a positive predictive value (PPV) of 53% and a negative predictive value (NPV) of 87% if GLS was less than −19.0% (p=0.011).11 Remarkably, neither radial nor circumferential strain were predictive for subsequent cardiotoxicity (p=0.25 and p=0.67, respectively) in that study. This particular finding was confirmed by Negishi's previously mentioned study.12

All these data taken together suggest that GLS could become the defining parameter of choice in the future redefinition of cardiotoxicity. This was illustrated by Thavendinarathan's paper in which he performed a variability analysis for LVEF with BPEF and 3DE. Thavendinarathan supposed that cardiotoxicity is improbable if GLS remains stable (no value above –16%) during the 1 year follow-up. This allowed him to attribute the variability of LVEF and volumes during follow-up to the LVEF calculation method only.6

Definite proof of the superiority of GLS in predicting cardiotoxicity should come from multicentre, prospective studies in large cohorts with long follow-ups (at least 5 years) and treated with chemotherapy regimens that are as homogeneous as possible. These studies are still lacking, mainly because chemotherapy is tailored for each individual patient according to body surface area (BSA) based dosing schemes, their haematologic tolerance, and the estimated risk of cardiotoxicity. Even then, drawing conclusions will remain difficult because of the partly genetic susceptibility for cardiotoxicity.15 Furthermore, confounding factors such as arterial hypertension, smoking, coronary disease, and diabetes might further spread out the time course of cardiotoxicity. Finally, the JUSTICE study13 evaluated the inter-vendor variability of GLS (as well as GRS and GCS) values as being pronounced, so that strain values obtained from different brands of machine during follow-up are not interchangeable.

Tissue Doppler imaging derived peak systolic mitral annular velocity (S’)

Peak systolic mitral annular velocity (S’) is another simple longitudinal deformation parameter with conflicting results in predicting cardiotoxicity. Even though reproducibility is not optimal due to its angle dependancyw26 and its age and gender variability,w27 Fallah-Rad et al showed that a significantly reduced S’, detected 3 months after combined anthracycline and trastuzumab treatment, predicted an LVEF decrease after 6 months with high sensitivity (93%) and specificity (99%).w20 In contrast, Appel et al w28 and Negishi et al12 could not find any longitudinal tissue Doppler imaging (TDI) parameter which predicted a future LVSD. Moreover, Ho et al showed that S’ can decrease in asymptomatic patients without future LVEF drop.10



A rise in troponin is considered the gold standard biomarker for myocardial injury from any cause, including chemotherapy.16 Its strengths are related to an almost absolute cardiac specificity, high sensitivity, wide diagnostic window, and a robust biochemical assay.w29

Lipshultz et al were the first to report a troponin T (TnT) rise in 30% of children treated with doxorubicin, with the magnitude of rise predicting LV dilatation and decreased wall thickness.w30 Furthermore, they proved that those subjects who demonstrated at least one TnT rise during chemotherapy showed significant late cardiac abnormalities at echocardiography.17

Cardinale et al confirmed these findings with troponin I (TnI) in an adult population receiving high dose anthracycline based chemotherapy. They proved that a rise in TnI could predict the very early development of LVSD and its severity: approximately half of the TnI rise occurred within 12 h of the start of chemotherapy, and the remainder occurred 12–72 h after the start. At 3 months, the decline in LVEF was significant and persisted until the end of follow-up.18 The same protocol was repeated in a much larger population (N=703) treated with less aggressive chemotherapy regimens for various malignancies, where patients were randomised according to three distinct patterns in TnI rise measured before, 3 days, and 1 month after completion of therapy. Patients devoid of a TnI rise (pattern 1) were considered as low risk for cardiotoxicity; they did not develop a significant decline in LVEF and the incidence of cardiac events was only 1%. In contrast, patients with a TnI rise developed a significant decrease in LVEF and a notably increased incidence of cardiac events. In particular, those patients with a persistent TnI rise 1 month after chemotherapy termination (pattern 3) had greater cardiac impairment and a higher incidence of major cardiac events, compared to those with only an early and temporary TnI rise (pattern 2) (84% vs 37%, p<0.001). Moreover, the correlation between maximal TnI value and maximal LVEF reduction improved notably from pattern 2 to 3 (from R=0.78; p<0.0001 for pattern 2, to R=0.92; p<0.0001 for pattern 3).19 Taking into consideration the very high NPV (99%) of TnI, it seems that patients with pattern 1 do not need close cardiac surveillance after chemotherapy.

Many studies have shown similar relations between troponin rise and subsequent decrease in LVEF.w31 In addition, TnI was found to be successful in differentiating reversible from irreversible myocardial injury during additional trastuzumab therapy after anthracycline administration: patients with a TnI rise were shown to have a 62% chance of LVSD, a higher incidence of major adverse cardiac events (MACE), and even a threefold decrease in the likelihood of recovery, compared to patients without a TnI rise (only 5% developed LVSD, p<0.001).20

In another prospective study, a TnT rise in 10% of sunitinib or sorafenib (tyrosine kinase inhibitors) treated patients for metastatic renal cell carcinoma was predictive (90%) of later LVSD.w32

Pooled data on troponin release confirm that: (1) it reflects early myocardial cell damage and death; and (2) a critical amount of myocardial damage is necessary before regional LVSD and more later LVEF drop occurs. Nevertheless, three practical concerns persist in the clinical use of troponin measurement for the early detection of cardiotoxicity. The first concern is that, as the rises in troponin in the different studies were detected at different time points in the chemotherapy regimens, the standardised time point at which the troponin measurement should be done has not yet been elucidated. The second concern is that, despite the high NPV of TnI during monitoring, the time point at which 100% specificity for no further troponin release would be reached has not yet been identified.w33 However, two groups of authors have applied a method with good results in clinical practice.11 w34 Interestingly, Sawaya et al found that only the combination of high sensitivity TnI (hsTnI) and GLS at 3 months after the start of chemotherapy was able to predict cardiotoxicity, while neither GLS nor TnI alone were able to do so.11 Third, and most importantly, most of the evidence for troponin monitoring has been collected from anthracycline treated populations, directly reflecting its cardiotoxic mechanism. As the cardiotoxic mechanism of trastuzumab is fundamentally different, and the literature reports much less cardiotoxicity after trastuzumab therapy in general, few data on troponin release are available in trastuzumab treated populations.20 Therefore, the evidence for serial monitoring of troponin rise during any cardiotoxic anticancer treatment cannot be applied universally yet.

Natriuretic peptides

Since the most feared manifestation of cardiotoxicity is CHF, it was not surprising that soon after discovery of natriuretic peptides, the first papers linking B-type natriuretic peptide (BNP) rise with cardiotoxic LVSD appeared in the literature.w35 Many studies followed patients of various age groups with differing malignancies who were receiving heterogeneous cardiotoxic treatment.

The prospective data for predicting cardiotoxicity through natriuretic peptide dosage are conflicting. Sandri et al21 published a prospective study in which 33% of patients with persisting high N-terminal pro BNP (NT-proBNP) values 72 h after chemotherapy administration developed a decrease in LVEF (from 62.8% to 45.6%, p<0.001) 1 year after treatment. Many reports are consistent with those findings, but a few did not find a correlation between a BNP rise and the development of LVSD.11 w36 w37 Explanations for these differences should be sought in small sample sizes, heterogeneity in populations, treatments and laboratory assays, and variability in follow-up time.w34 Prospective multicentre studies with large populations, standardised dosage methods, well defined sampling time, and cardiac end points are still needed to evaluate further the capability of natriuretic peptides in predicting cardiotoxicity.

Monitoring cardiac toxicity of chemotherapy: key points 2

  • Global longitudinal strain (GLS) with speckle tracking has proven to be a better prognosticator for global cardiac death and morbidity compared to LVEF.

  • A relative 10% reduction in GLS has repeatedly proven to predict later significant LVEF decrease.

  • A rise in troponin is the gold standard biomarker for myocardial injury from any cause.

  • Early troponin I (TnI) rises during chemotherapy predict later cardiac events well, especially congestive heart failure.

  • The negative predictive value of a TnI rise is 99%.

  • There is no available evidence for chronic troponin monitoring after treatment.

  • There is no evidence yet supporting the monitoring of natriuretic peptides for cardiotoxicity detection.

  • Combination of early TnI rise and GLS reduction might redefine cardiotoxicity criteria in the near future.

  • This should be tested in large prospective and multicentred studies with long follow-up.

Recent pathophysiological insights might open new diagnostic pathways

Newer biomarkers

Enhanced releases of fatty acid binding proteinw38 and glycogen phosphorylase isoenzyme BB,w39 newer biomarkers of myocardial ischaemia and necrosis, as well as of cytokines,w40 biomarkers of inflammation, have all been associated with anthracycline administration. However, all data are still preliminary or fragmentary and need further investigation/confirmation. In addition, biomarkers of endothelial dysfunction have been shown to increase after chemotherapy,w41 w42 but no correlation with long term CV events has been demonstrated, so their predictive value for cardiotoxicity has yet to be defined.

CMR beyond LVEF: delayed enhancement

A pilot study in 10 anthracycline and trastuzumab (Herceptin) treated breast cancer patients with already diagnosed cardiotoxicity (systolic dysfunction) showed a striking pattern of mid myocardial delayed enhancement,w43 comparable with the presenting image seen in myocarditis.

In another pilot study, myocardial T1 mapping identified, in some patients, significantly increased myocardial signal intensity within 3 days after the first anthracycline dose, followed by a significant decrease in LVEF on day 28 after the start of chemotherapy.w44 Larger cohorts could potentially reveal a clear threshold beyond which the risk of developing LVSD increases. If those data could be correlated with cumulative chemotherapy doses in larger scale cohorts, it might impact heavily on cardio-oncological prevention strategies. Patients showing increased T1 weighted myocardial signal intensity at relatively low doses seem to be at high risk for developing early cardiotoxicity, while those patients not showing increased signal intensity might tolerate more aggressive chemotherapeutic regimens.w45

MUGA beyond LVEF

MUGA specific diastolic parameters have been developed and evaluated: peak filling rate (PFR) and time to peak filling rate (TPFR). But as with echocardiographic diastolic parameters, they are age dependent, complicating correct clinical interpretation.w46 Additionally, in a study comparing PFR and TPFR with LVEF, both decreased simultaneously instead of the expected delay in LVEF drop.w47

123I*-MIBG, an analogue of norepinephrine, has been tested in the setting of high dose anthracycline regimens. In analogy with patients with CHF due to any cause, the ratio of heart to mediastinal MIBG uptake (H/P ratio) is decreased,w48 and seems to correlate inversely with cumulative anthracycline doses.w49 Moreover, the decreased H/P ratio precedes decline in LVEF.w50 The clinical relevance has still to be determined since we do not know which H/P ratio slope curve portends excessive risk of future LVSD, and if the method also works for patients receiving lower cumulative doses.w45

111Indium-labelled trastuzumab has been used to visualise metastatic breast cancer expressing HER-2/Neu positivity before treatment.w51 In one study, 86% of the patients demonstrating 111I-trastuzumab uptake (35%) developed CHF and good oncological response to treatment (including trastuzumab). The latter 65% with no tracer uptake did not develop cardiac events. Unfortunately, tracer uptake decreases after trastuzumab treatment in HER-2/Neu positive breast cancer, making the predictive value of tracer uptake on the development of heart failure poor.w52 In conclusion, dedicated studies are needed which attempt to correlate baseline HER-2 expression levels, anthracycline and trastuzumab regimens, and early/late cardiac side effects.w45

A new and integrated approach to detect preclinical cardiotoxicity in clinical routine

Available guidelines/methods for predicting cardiotoxicity after anthracycline treatment, based on LVEF alone, have been found to be inadequate.22 The clinical case shown in figure 1 illustrates this quite well. Meanwhile, additional anticancer treatments with (potentially) synergistic cardiotoxicity are used routinely.

Figure 1

Panels A and B: Multimodality LVEF and GLS at baseline (line A) and 1 year after start of treatment (panel B) in a breast cancer patient with AHT, DM and smoking, measured with (1) BPEF, (2) 3DE, (3) CMR , and (4) GLS with 2D strain. LVEF is unchanged with CMR and 3DE after 1 year, ‘excluding’ cardiotoxicity according to clinical practice guidelines. The BPEF drop is explained by its measurement variability. GLS drops by 5% over 1 year, with patchy reductions in segmental strain compared to very homogeneous segmental strain baseline, showing potential cardiotoxicity in development. Panel C: Evolution of BPEF, GLS, and NT-proBNP in the same patient the first year after the start of chemotherapy. At baseline, NT-proBNP is slightly increased, probably due to known AHT. Just after completion of chemotherapy, a clear bump appears in the time curve of NT-proBNP measurements. Simultaneously, GLS decreases accordingly, and reaches a stable but reduced level from 6 months, compared to baseline. In contrast, BPEF is unchanged except for temporal variability of BPEF. AHT, arterial hypertension; BPEF, biplane ejection fraction; CMR, cardiac magnetic resonance; 3DE, 3D echocardiography; DM, diabetes mellitus; GLS, global longitudinal strain (speckle tracking); LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro B-type natriuretic peptide.

Serial LVEF monitoring with BPEF alone is not accurate enough. The literature reveals that 3DE derived LVEF approaches best the gold standard MRI derived values, but does not resolve the limitation of LVEF being a late and insensitive prognostic marker of cardiotoxicity. This almost excludes the window for cardiotoxicity prevention.

2D GLS adds earlier and incremental prognostic information to LVEF, especially in those patients with (nearly) normal LVEF, but large studies with long follow-ups should be undertaken to confirm the still fragmented data. Furthermore, it should be stressed that serial 2D GLS values cannot be compared during follow-up if acquired with different machines.

Additionally, troponin rise seems an early, sensitive and reproducible biomarker predicting cardiotoxicity LVSD, but most of the data are derived from anthracycline treated patients only. Furthermore, we still need to determine when troponin should be measured during the anticancer treatment, to either confirm or exclude cardiotoxicity. Finally, there is no evidence on the need for chronic monitoring of troponins after treatment.

As a practical clinical tool, table 1 summarises the different diagnostic methods discussed above, mentioning statistical performance, study type, number of included patients from key reference papers, and finally strengths and weaknesses.

Table 1

Overview of different diagnostic methods for cardiotoxicity with statistical performance, reference paper(s), study type and number of included subjects, strengths and weaknesses

As explained above, it is too soon yet to impose a new diagnostic algorithm for the detection of cardiotoxicity. Early data with short follow-up suggest that the future approach will combine the monitoring of troponin release as the ultimate parameter of myocardial damage and GLS as the ultimate parameter of functional impairment in an attempt to enhance the statistical performance in predicting cardiotoxicity. Such a potential algorithm or screening protocol is proposed in figure 2.

Figure 2

Proposed algorithm for the detection and monitoring of cardiotoxicity based on the hypothesis that troponin I (Trop I), global longitudinal strain with speckle tracking (GLS), and potentially B-type natriuretic peptides (BNP) are accepted as diagnostic standards. Panel A shows a classic chemotherapy time line. After the initial baseline evaluation, the anthracycline based chemotherapy period starts, ending at 3 months usually, eventually followed trastuzumab (Herceptin) or taxanes. During the anthracycline treatment, troponin I measurements within 24 h of the start of each cycle until the end of anthracycline therapy should be performed. At the end of this phase, echocardiographic and BNP monitoring during eventual trastuzumab or taxane treatment can be started. Panel B illustrates the differentiation (after the baseline evaluation) between echogeneous patients, further followed with echo for LVEF and GLS, and non-echogeneous patients, followed with cardiac magnetic resonance (CMR) as fist alternative. In case CMR is not available, multiple gated acquisition (MUGA) scanning or finally contrast use for left ventricular (LV) opacification and biplane LVEF calculation can be used as a third alternative, depending on local availability or expertise (see * in panels B and C). Panel C shows follow-up for patients, either with echo or CMR. In the case of abnormal baseline evaluation with either method, a non-toxic or less toxic chemotherapy should be considered, as well as starting ACE inhibition (ACE-I) or β-blockade. Finally, the patient will need intensive follow-up. In the case of normal baseline screening with all methods, decision making depends on whether there is a troponin I rise during therapy or not. The first detected troponin I rise should bring the patient into the intensive follow-up stream with frequent LV function imaging and continued troponin I and BNP monitoring. Additionally, starting ACE-I or β-blockade should be considered, as well as chemotherapy change (type, dose, frequency, sequence in case of combination). If a troponin I rise is absent during therapy, a repeat evaluation 1 year after the start of therapy is sufficient. At or after 3 months, during potential trastuzumab or taxane treatment, only trastuzumab withdrawal can be considered in the case of persisting troponin I rise, relative 10% decrease in GLS, or absolute 10% decrease of LVEF. If not started yet, ACE-I or β-blockade should be started, and trastuzumab can be rechallenged eventually after, for example, 3 months with good prognosis (no further decrease in LV function).

Confirmation in large, multicentre prospective studies in homogeneous populations is necessary. Additionally, the follow-up should be extended to at least 5 years, because late cardiotoxicity has an equally poor prognosis as earlier cardiotoxicity. Since the clinical question of how to improve the prognosis of the cancer patient while at the same time limiting CV complications is pertinent and needs urgent answers, endorsement by international oncologic and cardiologic societies seems essential for the initiation and success of such studies.

You can get CPD/CME credits for Education in Heart

Education in Heart articles are accredited by both the UK Royal College of Physicians (London) and the European Board for Accreditation in Cardiology—you need to answer the accompanying multiple choice questions (MCQs). To access the questions, click on BMJ Learning: Take this module on BMJ Learning from the content box at the top right and bottom left of the online article. For more information please go to:

▸ RCP credits: Log your activity in your CPD diary online (—pass mark is 80%.

▸ EBAC credits: Print out and retain the BMJ Learning certificate once you have completed the MCQs—pass mark is 60%. EBAC/ EACCME Credits can now be converted to AMA PRA Category 1 CME Credits and are recognised by all National Accreditation Authorities in Europe (

Please note: The MCQs are hosted on BMJ Learning—the best available learning website for medical professionals from the BMJ Group. If prompted, subscribers must sign into Heart with their journal's username and password. All users must also complete a one-time registration on BMJ Learning and subsequently log in (with a BMJ Learning username and password) on every visit.


Special thanks to Benjamin Scott, MD, ZNA Antwerpen Belgium, for his critical review and for improving the clarity and readability of the paper.


  1. Statistical landmark paper evaluating intra-, inter- and test–retest variability of the biplane LVEF calculation method in normal and cardiac diseased subjects.
  2. Offers prospective proof that serial monitoring of LVEF with real-time 3D echocardiography approaches reproducibility of MUGA scans, compared to CMR as gold standard, and in the specific setting of chemotherapy treated patients.
  3. Landmark paper stating that cardiotoxicity is excluded if GLS stays stable during and after chemotherapy. Only available paper extensively and prospectively comparing within one cohort variability of 2D and 3D LVEF, with and without contrast. Comprehensively explains why automated endocardial border tracking does improve reproducibility of LVEF calculation, and contrast (by excluding the former) does not.
  4. Concise but systematic and practical review of all available methods for detection of cardiotoxicity, outlining their level of evidence, advantages and drawbacks. It is outdated on the evidence for the use of deformation imaging, especially GLS, in detecting cardiotoxicity.
  5. Illustrates the huge diagnostic power of GLS in being able to distinguish smoking from non-smoking chemo treated patients on the basis of GLS changes after treatment, compared to the control group.
  6. First paper proving that combining GLS assessment and ultrasensitive troponin I rise can predict cardiotoxicity and might be the way to further improve prediction of cardiotoxicity and apply cardio-oncological prevention.
  7. Confirms that GLS reduction is an independent, and the strongest, predictor (over LVEF) of subsequent cardiotoxicity after chemotherapy in a larger cohort (N=81) than in most papers. Confirms that monitoring strain changes is more accurate than monitoring absolute strain values, generating an optimal cut-off in GLS reduction of 11% for the prediction of cardiotoxicity.
  8. Landmark paper for variability analysis and normal values in 2D strain analysis. Highlights that 2D strain data are not interchangeable for longitudinal follow-up or cross-sectional assessment of LV function due to important inter-vendor variability.
  9. Extensive but comprehensive key paper reviewing the use of biomarkers, especially troponins, in cardio-oncology, ranging from their diagnostic performance and prognostic abilities to their potential guiding of cardio-oncological prevention therapy.
  10. Landmark paper, illustrating that the very early TnI rise during chemotherapy is able to predict a decline in LVEF.
  11. Key paper for the potential use of TnI rise in clinical routine, on the basis of its high NPV in an anthracycline-only treated population.
View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors All authors contributed to the following below: (1) substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data; (2) drafting the article or revising it critically for important intellectual content; and (3) final approval of the version to be published.

  • Competing interests In compliance with EBAC/EACCME guidelines, all authors participating in Education in Heart have disclosed potential conflicts of interest that might cause a bias in the article. The authors have no competing interests.

  • Provenance and peer review Commissioned; externally peer reviewed.