To analyze the utility of cardiotoxicity monitoring, focusing on the impact of imaging-guided interventions, a comprehensive set of clinical data was required. Retrospective data including the clinical course of 109 patients from the first visit to the end of one year of treatment were collected from the cardio-oncology clinic of Hospital Santa Maria between November 2015 and November 2016. Patients in this study followed individual anti-cancer therapies and cardiac medication according to their cancer type. The patients’ mean age was 66 years. The mean age of the 51 patients with breast cancer was 63 years. A summary of the dataset is shown in Table 1.

This decision-analytic model assessed the morbidity, mortality, and costs associated with two strategies: the current standard strategy of initiating cardioprotective medications after diagnosis of EF-CTX (defined as an asymptomatic decline in LVEF of ≥10% to a final value ≤55%) or symptomatic heart failure; and a strategy of universal cardioprotection (UCP) for all patients at the time of chemotherapy. Cardioprotection was defined as concurrent enalapril and carvedilol up-titrated to their maximum dose, as used in the active treatment arm of a large recent randomized controlled trial,8 and used throughout the five years of modeling specialty. This Markov model used Monte Carlo simulations (Visual Basic for Applications in Excel) to assess the clinical and economic consequences of alternative cardioprotective strategies in a hypothetical cohort of 1000 patients in a microsimulation model.9 Beta distributions were assigned to probabilities and utilities, and gamma distributions to costs10 based on standard errors derived from our dataset and the literature. Cost-effectiveness acceptability curves (CEACs) were used to report the probability of incremental cost-effectiveness ratios (ICERs) at the willingness-to-pay threshold. The study was performed in accordance with the Consolidated Health Economic Evaluation Reporting Standards, as detailed by the International Society for Pharmacoeconomics and Outcomes Research.

Following Nolan et al.,11 the costs and QALYs of the treatments were estimated over a five-year period of the cohort, because transition probabilities beyond this period are not currently well described. It was assumed that all interventions took place at the start of the time horizon, and all future costs and benefits were discounted by 3% per annum. Given that probability data from the literature use a one-year cycle length, we transformed our three-monthly transition probabilities (in which the cycle length is four months) to a one-year cycle length.

The Markov model accounted for the dynamics of cardiac screening and use of cardioprotective medications in a cohort of 1000 patients scheduled to receive chemotherapy for cancer. The base case is a 63-year-old woman with breast cancer receiving sequential anthracycline and trastuzumab therapy, which was applied to all cohort patients. This ‘average’ patient was assigned to one of the two treatment strategies. These patients progressed through the Markov model (Figure 1) on the basis of transition probabilities (Table 2), informed by utilities (Table 3) and costs (Table 4).

Figure 1.

Transition states incorporated into the Markov model. In the Markov model patients begin in a state of no cardiac symptoms (i.e. well). During the next cycle, patients can remain in that state, die or be diagnosed as having cardiotoxicity or symptomatic heart failure. Patients can subsequently either remain in those states or die. Patients with cardiotoxicity may also be diagnosed with symptomatic heart failure before dying. Patients taking cardioprotective medications who can have side effects and/or discontinue the medication but do not leave the health state are represented by the oval.

(0.11MB).

In the Markov model patients begin in a state of no cardiac symptoms (i.e. well). During the next cycle, patients can remain in that state, die or be diagnosed as having cardiotoxicity or symptomatic heart failure. Patients can subsequently either remain in those states or die. Patients with cardiotoxicity may also be diagnosed with symptomatic heart failure before dying. Patients taking cardioprotective medications may suffer from side effects and/or discontinue medication but do not leave the health state; these are represented by the oval in Figure 1.

Data on transitions between health states were obtained from the dataset, the literature and expert sources (Table 2). From our dataset, transition probabilities were calculated from the proportion of patients in each state, following the procedure suggested by Black et al.12 and Miller and Homan.13 We closely followed Nolan et al.,11 who performed a similar cost-effectiveness analysis and gathered probabilities, cost and utilities from the literature. Transition probabilities were separated into two categories, initial year of treatment and subsequent years.

The economic analysis was from the healthcare payer's perspective and therefore used the amount reimbursed to the provider as the cost of care. Information regarding costs was obtained from our dataset and the literature (Table 3).

LVEF-guided echocardiographic screening costs were calculated by allocating the Portuguese official healthcare charges14 to the number of echocardiographic studies observed in the 109 patients. Calculation of medication costs included the medication indicated in the data, the active substance and dose, the price from the Portuguese National Authority for Medicines and Health Products (INFARMED) reference prices for generic medication,15 the government reimbursement rate and the number of packages needed for one year of treatment. Costs were expressed in 2019 euros and a willingness-to-pay threshold of €22 986 per QALY was applied, as this represents the 2019 average annual gross domestic product (GDP) of Portugal16 and a national per capita GDP value has been suggested by the World Health Organization to represent how much a nation can reasonably spend to save each QALY.17 Costs published in the literature more than two years earlier were corrected for intervening currency fluctuations using the Banco de Portugal's currency converter.18

Table 4 lists the utilities used in this study (values obtained from preferences associated with health-related quality of life where fully healthy=1.0 and dead=0.0). We followed Nolan et al.11 and the values were obtained from Gohler et al.,20 Lewis et al.,21 and Imperiale et al.22 Gohler et al.20 studied utility estimates for decision-analytic modeling in heart failure, with health states based on New York Heart Association classes and number of rehospitalizations. Lewis et al.21 assessed preferences for quality of life or survival expressed by patients with heart failure using the Minnesota Living with Heart Failure questionnaire. The patients’ utility for medication side effects also refers to patients with heart failure.

To determine the applicability to other types of cancer of the cost-effectiveness analysis of cardioprotective strategies, one additional scenario was run using the same Markov model. The scenario used the mean of all patients in our database, i.e. a 66-year-old patient with any type of cancer that involved chemotherapy treatment. This scenario includes patients with common cancer types, such as breast cancer, colon and rectal cancer, gastric cancer, kidney cancer, leukemia, lung cancer, melanoma, non-Hodgkin lymphoma and prostate cancer. Transition probabilities shown in Table 2 reflect the higher incidences of cardiovascular complications due to more aggressive chemotherapeutic regimens for cancers like Hodgkin lymphoma and acute myeloid leukemia.

Probabilistic sensitivity analysis was performed by varying all variables simultaneously over their plausible ranges. This process involved assigning distributions to 25 variables (15 event probabilities, six costs, and four utilities) used in the model. For the Monte Carlo simulation, 1000 iterations were run for both strategies using unique combinations of the 25 distributions, with values for each variable selected randomly. The differences between the strategies for each run were recorded to determine how often each was considered more cost-effective.

In probabilistic sensitivity analysis, a range of values reported from our dataset and found in the literature for transition probabilities, costs and utilities, were considered. When data were available, low and high values were chosen to reflect ranges in the literature. The model structure was based in part on other models in the literature, and was reviewed by clinicians involved in the care of cancer patients and chemotherapy-related cardiotoxicity.

In the reference case of a 66-year-old patient taking anthracycline and trastuzumab for breast cancer, the outcomes of an LVEF-guided strategy (4.22 QALYs, €2594 over five years) were superior to those of UCP (3.42 QALYs, €3758 over five years). For a 66-year-old patient with any type of cancer, the outcomes of an LVEF-guided strategy (3.76 QALYs, €6718 over five years) were also superior to those of a UCP strategy (2.97 QALYs, €7995 over five years).

In Table 5, it can be seen that the findings for all patients are broadly similar to those of the base case, with the ICER for LVEF-guided strategy being negative for both, around 1.5 and 1.6, i.e. an LVEF-guided strategy is less expensive and more efficacious than UCP. However, the QALYs gained are smaller and the costs higher for all cancer patients compared to breast cancer patients, under both strategies.

The effects of parameter uncertainty on the frequency of transition probabilities, utilities and costs were examined using a Monte Carlo analysis with 1000 simulations. This demonstrated that LVEF guidance was the optimum strategy for breast cancer patients, with a mean survival of 3.96 QALYS vs. 3.17 years with UCP (Table 6).

Figures 2 and 3 show the cost-effectiveness acceptability curves representing the probability that an LVEF-guided strategy is cost-effective for a given maximum willingness-to-pay threshold per QALY gained. The graphs are based on 1000 Monte Carlo simulations, drawing parameters for each input from probability distributions. Both cost-effectiveness acceptability curves demonstrated that an LVEF-guided strategy was consistently cost-effective across a broad plausible range of willingness-to-pay thresholds (from €0 to €50 000 per annum).

Figure 2.

Cost-effectiveness acceptability curve for breast cancer patients representing the probability that each treatment strategy is cost-effective for a given maximum willingness-to-pay threshold per quality-adjusted life-year gained. The graph is based on 1000 Monte Carlo simulations, taking parameters for each input from probability distributions.

(0.1MB).

Figure 3.

Cost-effectiveness acceptability curve for all cancer patients representing the probability that each treatment strategy is cost-effective for a given maximum willingness-to-pay threshold per quality-adjusted life-year gained. The graph is based on 1000 Monte Carlo simulations, taking parameters for each input from probability distributions.

(0.09MB).

Incremental cost-effectiveness scatter plots of results of 1000 pairs of differences in costs and QALYs (Figures 4 and 5) show that an LVEF-guided strategy is more cost-effective than UCP. The comparison between LVEF and UCP shows clear benefit for LVEF for breast cancer patients, with LVEF being more cost-effective in 78.2% of instances, and for all patients, with LVEF being more cost-effective in 75% of instances.

Figure 4.

Incremental cost-effectiveness bootstrap scatter plot for breast cancer patients. The y-axis represents the difference in mean costs (€2019), and the x-axis represents the difference in mean quality-adjusted life-years (QALYs). In 78.2% of iterations a left ventricular ejection fraction (LVEF)-guided strategy provided additional QALYs. Iterations in which an LVEF-guided strategy was less expensive and more efficacious than universal cardioprotection (UCP) accounted for 42.0% of the 1000 iterations. Iterations for which an LVEF-guided strategy provides additional QALYs at additional cost accounted for 36.2%. Iterations for which an LVEF-guided strategy is more expensive and less efficacious than UCP accounted for 15.9%, and iterations for which an LVEF-guided strategy is less expensive and less efficacious than the comparator accounted 5.9%.

(0.11MB).

Figure 5.

Incremental cost-effectiveness bootstrap scatter plot for all cancer patients. The y-axis represents the difference in mean costs (€2019), and the x-axis represents the difference in mean quality-adjusted life-years (QALYs). In 75.0% of iterations a left ventricular ejection fraction (LVEF)-guided strategy provided additional QALYs. Iterations in which an LVEF-guided strategy was less expensive and more efficacious than universal cardioprotection (UCP) accounted for 39.8% of the 1000 iterations. Iterations for which an LVEF-guided strategy provides additional QALYs at additional cost accounted for 35.2%. Iterations for which an LVEF-guided strategy is more expensive and less efficacious than UCP accounted for 17.3%, and iterations for which an LVEF-guided strategy is less expensive and less efficacious than the comparator accounted 7.7%.

(0.12MB).