Table 2 Predictive models for cardiovascular events in cancer patients

Ezaz et al. [15]

From SEER database: N = 1664 pts. treated for HER2+ BC with systemic therapies

1 point: past medical history of hypertension, diabetes, or age 75–79 years.

2 points: history of coronary artery disease, renal failure, or atrial fibrillation or flutter, having received any chemotherapy or > 80 years of age

Three risk groups: low (0 to 3 points), medium (4 to 5 points), and high (≥6 points) risk strata with 3-year CE rates of 16.2, 26.0, and 39.5%, respectively

Training set of 70%, internal validation in the other 30% with strong performance of the model

Validation in additional external cohorts.

No information on cardiovascular medication use

Rushton et al. [16]

N = 143 patients with HER2+ BC referred to a cardio-oncology clinic at a tertiary care center

Sensitivity analysis to validate model composed by Ezaz et a [16]

Low risk: 42% CE rate, 13% permanent HF

Moderate risk: 64% CE rate, 14% permanent HF

High risk: 30% CE rate, 20% permanent HF

Low cardiac risk score had a negative predictive value of 94% for permanent cardiotoxicity.

Highly selected population.

Sub-optimal performance in high risk group.

Fogarassy et al. [8]

Nationwide health care databases, N = 8068 BC pts. treated with epirubicin

Risk-prediction score for HF composed of age, diabetes mellitus, hypertension, coronary artery disease, stroke, epirubicin dose, docetaxel dose, capecitabine, gemcitabine, bevacizumab and cancer stage

Five score point categories and corresponding risk for HF; score 1–7 HF 2.1%, score 8–9 HF 5.0%, score 10–12 HF 10.3%, score 13–18 HF 22.1%, score 19–26 HF 31.7%

Large dataset, Training set of 70%, internal validation in the other 30%

Information on cardiovascular medication use.

External validation.

Romond et al. [17]

Analysis from NSABP B-31 trial, N = 1830 pts. with HER2+ BC

Retrospective regression analysis to reveal predictors for cardiac events: formula to calculate cardiac risk score

Cardiac risk score based on age and baseline LVEF by MUGA

High discriminate ability (C-index 72%) in associating the length of time to a cardiac event with the probability of not experiencing CEs.

No external validation of the risk score

Hermann et al. [18]

Literature- and expert-based recommendation

Type of treatment, age, gender, history of CVD or presence of risk factors for CVD

No statistical validation

Easily accessible variables

Test algorithm in patient population for clinically relevant endpoints.

Abdel-Qadir et al. [19]

Real-world population EBC Ontario 2003–2015, N = 90,104 (2/3 training, 1/3 validation set)

Risk-prediction score for MACE composed of age, hypertension, diabetes, ischaemic heart disease, atrial fibrillation, HF, cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease, and chronic kidney disease

Ten-year MACE incidence was > 40-fold higher for patients in the highest score decile compared to the lowest. The c-index was 81.9% (95% confidence interval 80.9–82.9%) at 5 years and 79.8% (78.8–80.8%) at 10 years in the validation cohort, with good agreement between predicted and observed MACE incidence.

Clinically relevant long-term outcome

No incorporation of cancer and treatment-related variables

Dranitsaris et al. [20]

Metastatic breast cancer pts. treated with anthracyclines (doxorubicin or liposomal doxorubicin), N = 509

Risk scoring algorithm (range 0–62) based on number of cumulative cycles, patient age and weight, previous anthracycline exposure and poor performance statu

A ROC analysis had an area under the curve (AUC) of 0.84 (95% CI: 0.79–0.89). A precycle risk score cutoff of ≥30 to < 40 was identified to optimally balance sensitivity (58.5%) and specificity (89.0%).

Easily accesible variables

Validation in external cohorts.

Add information on risk factors for CVD and medication use .