Morning administration of anthracyclines is associated with a lower risk of cancer therapy-related cardiac dysfunction

Abstract

Aims

Pre-clinical studies point towards an administration time-dependency of anthracycline-induced cancer therapy-related cardiac dysfunction (CTRCD). This retrospective study aimed to investigate the association between time-of-day of AC administration and CTRCD.

Methods and results

Patients from two cardio-oncology outpatient clinics, treated with ACs for any malignancy, were included. Percentage of afternoon AC administration was calculated: cumulative dose administered in the afternoon (12 p.m.–11:59 p.m.)/total cumulative dose. Three groups were defined: morning group ≥ 50% of ACs in the morning (12 a.m.–11:59 a.m.), afternoon group ≥ 50% of ACs in the afternoon, and intermediate group = exactly 50% of ACs in the morning and afternoon. Associations between AC timing and occurrence of CTRCD and heart failure (HF) were assessed using survival analyses. Of 270 included patients, 66 developed CTRCD and 17 developed HF. Compared with the morning group, the afternoon group had a higher risk of developing CTRCD: hazard ratio (HR) 2.88 (95% CI: 1.52–5.44). When considering percentage of ACs administered in the afternoon as a continuous variable, the HR for developing CTRCD was 1.14 (95% CI: 1.04–1.24) for each subsequent 10% of afternoon administration. Results were consistent across sensitivity analyses of age, sex, body mass index, malignancy type, cumulative AC dose, and HFA-ICOS risk score. Congruently, the continuous variable of afternoon AC administration was associated with higher risk of HF: HR = 1.19 (95% CI: 1.01–1.41).

Conclusion

Afternoon administration of ACs is associated with an increased risk of developing CTRCD and HF, suggesting that morning administration may be preferred. Before widespread implementation, these findings should be confirmed in an RCT.

Graphical Abstract

Graphical Abstract.

Summary of study design and main findings. 95% CI, 95% confidence interval; CTRCD, Cancer therapy-related cardiac dysfunction; HF, Heart failure; HR, Hazard ratio.

Introduction

Anthracyclines (ACs) are chemotherapeutics that form a key component in the treatment of many cancers, including haematological malignancies and breast cancer. Although their antineoplastic properties greatly improve oncological survival, treatment with ACs increases the risk of cardiovascular disease. Their cardiotoxic properties can cause permanent cardiac damage, which can manifest as cancer therapy-related cardiac dysfunction (CTRCD).¹ While usually asymptomatic in the early stages, CTRCD can progress into heart failure (HF), resulting in poor prognosis and quality of life in cancer survivors.² Even though the development of CTRCD has been linked to cumulative dose, and consequently maximum doses have been limited, AC-induced CTRCD still has an overall incidence of at least 9% and strongly depends on patients’ risk profiles.^3–5 In recent years, preventive strategies have been investigated, resulting in the approval of dexrazoxane and liposomal AC formulations in selected patients.¹ As cancer incidence is rising and ACs continue to be the standard of care in many cancers, AC cardiotoxicity remains a relevant issue for which widely applicable preventive strategies are warranted.⁶

Chronomodulated chemotherapy is a promising approach that could limit CTRCD incidence. A systematic review performed by our research group showed that this strategy, which encompasses the adjustment of chemotherapeutic administration time to the body’s circadian rhythms (i.e. 24-h rhythms), has the potential to reduce toxicity, while maintaining oncological efficacy.⁷ Unfortunately, in these randomized controlled trials (RCTs) assessment of the effect of administration time on AC cardiotoxicity was very limited. However, chronomodulated chemotherapy for AC cardiotoxicity prevention has shown favourable results in preclinical studies. In vitro experiments have led to the detection of a time-dependent damage response to doxorubicin in various cardiomyocyte models.^8,9 Similar findings were obtained in an animal study, where specific administration times of doxorubicin led to a decrease in cardiotoxicity in rodents.¹⁰ Hence, there is promising evidence that chronomodulated administration of ACs might reduce the occurrence of CTRCD.

Circadian rhythms are foundational to the mechanism of action of chronomodulated chemotherapy. They encompass the intrinsic 24-h rhythms present in almost all cells, which get entrained to the environment by external factors. These rhythms lead to diurnal (day/night) fluctuations in many processes, such as cell metabolism, DNA damage response, and immune responses.^11,12 Additionally, circadian rhythms can affect pharmacokinetics, e.g. through circadian rhythmicity in liver and kidney function.¹³ On a molecular level, circadian rhythms consist of a series of transcriptional-translational feedback loops, which have been found in every cardiovascular cell type.¹⁴ Congruently, cardiovascular (patho-)physiology is known to be greatly influenced by the circadian clock.¹⁴ Although the pathways underlying AC cardiotoxicity have not yet been elucidated fully, proposed mechanisms include mitochondrial dysfunction, excessive production of reactive oxygen species, direct DNA damage through intercalation, interference with topoisomerase 2B, and calcium overloading.¹⁵ Given the broad influence of circadian rhythms on intracellular processes in the heart,¹⁴ many of these mechanisms could be subject to 24-h rhythms, which may lead to a diurnal pattern in AC-induced myocardial damage.

To date, clinical studies investigating the association between AC administration time and CTRCD are lacking. In this study, we aimed to examine this association using real-world data from two cardio-oncology outpatient clinics in the Netherlands.

Methods

Study design and participants

This retrospective observational cohort study included electronic health record (EHR) data from patients aged ≥18 years, who visited the cardio-oncology outpatient clinics in the University Medical Centre Utrecht (UMCU) in the Netherlands from April 2015 to April 2023, and in the St. Antonius Hospital Nieuwegein (AZN) from March 2019 to October 2020. Initially, existing EHR data was collected, and in accordance with the system of no objection, informed consent was not mandatory, as outlined in Recital 62 from the European General Data Protection Regulation (EU GDPR).¹⁶ In addition, the local Medical Research Ethics Committee declared that the Medical Research Involving Human Subjects Act (WMO) does not apply to this part of the data collection. From 1 March 2019 onward, the Dutch cardio-oncology registry (ONCOR) initiated, for which all included patients signed informed consent forms. No methodological differences are present between both methods of data collection, since in both cases this encompassed the collection of the same datapoints from the EHR in the same population. This study conforms to the principles outlined in the Declaration of Helsinki. For an extensive description of ONCOR, we refer to the design paper.¹⁷

Referral to cardio-oncology outpatient clinics is typically reserved for patients at an increased risk of developing CTRCD. At the included centres, this risk was either calculated by the referring physician using the cardiotoxicity risk score (CRS), or a risk assessment was made based on the physician’s clinical evaluation.^18,19 In addition, patients with an indication for allogenic stem cell transplantation were referred after initial treatment with ACs. The CRS is based on a combination of medication-related and patient-related risk factors. In practice, this score was modified, attributing four risk points to a cumulative dose of ACs >240 mg/m² and two points to a cumulative dose of ≤240 mg/m². Patients treated with ACs were included if at least 75% of administration time points could be ascertained. Patients with all types and stages of cancers were included, and ACs were usually part of a multi-drug systemic anti-cancer treatment. Exclusion reasons were cumulative AC dose <50 mg/m², 24-h infusions of ACs, treatment with liposomal AC formulations, and the absence of outcome measures. Patient data were stored in research data platform REDCap, hosted by the Netherlands Heart Institute.¹⁷ All outcome data were updated on 1 October 2023.

Outcomes

The primary outcome was AC-induced CTRCD, which was defined as a left ventricular ejection fraction (LVEF) drop of at least 10% points compared with baseline (i.e. before start of AC treatment), to a value below 53%.²⁰ In the absence of a baseline LVEF measurement, an LVEF below 50% after start of AC treatment was considered CTRCD.²⁰ In both definitions, the reduction in LVEF had to be attributed to AC cardiotoxicity and alternative aetiologies had to be considered unlikely based on clinical findings and additional diagnostic testing such as CMR and CT-scan, where appropriate. The secondary outcome was HF, which involved both the diagnosis CTRCD, combined with an LVEF nadir value below 40%, or HF symptoms. Unclarity in the EHR with regards to scoring of events was resolved through expert consensus (MIP and AJT). The frequency of cardiovascular monitoring was dependent on CRS, and additional imaging was performed when HF was clinically suspected.¹⁸ As a safety measure, mortality outcome data was collected from the EHR and descriptive statistics were reported. No statistical testing was performed for the potential association between AC administration time and survival, since oncological prognosis was heterogeneous, as patients with varying malignancy types and disease stages were included. For all outcomes, follow-up time was determined as the difference in time between the first AC exposure and the event or last follow-up, whichever occurred first.

Exposures

The mean AC administration time was calculated ((end time–start time)/2) for each cycle. The day was divided into two sections: 12 a.m.–11:59 a.m. and 12 p.m.–11:59 p.m., labelled ‘morning’ and ‘afternoon’, respectively, because most treatments were administered during daytime working hours. Based on the mean administration time, the percentage of cumulative AC dose received in the afternoon was calculated for each patient, adjusting for dose per cycle. For patients who received different types of ACs, cumulative dose was calculated using the following equivalence ratios: doxorubicin = 1, epirubicin = 0.8, daunorubicin = 0.6, idarubicine = 4.5, mitoxantrone = 10.5.²¹ Patients were divided into three groups based on the timing of AC administration: morning group ≥ 50% in the morning, afternoon group ≥ 50% in the afternoon, and intermediate group = equal distribution of morning and afternoon administration.

Potential confounders and effect modifiers

All relevant data was collected given it could influence AC cardiotoxicity occurrence and/or circadian rhythms. The most important factors that may modify cardiotoxicity risk include cumulative AC dose, AC infusion duration, sex, age, body mass index (BMI), malignancy type and concomitant oncological treatments, such as trastuzumab and thoracic radiotherapy, cardiovascular comorbidities, and cardiovascular medication. Additionally, socioeconomic status (SES) was considered. As for circadian rhythms, data on treatment with systemic corticosteroids, benzodiazepines and melatonin were considered relevant as these may influence the biological clock.

A directed acyclic graph was created to determine potential confounders and effect modifiers. The following potential confounders were considered: age, biological sex, malignancy type, cumulative AC dose, mean infusion duration, SES, and trastuzumab treatment.^4,5,22,23 Malignancy type consisted of haematological malignancy, breast cancer, or other malignancy. As an additional confounder, SES was approximated using postal codes, which were linked to published data from Statistics Netherlands, from 2021.²⁴ SES data was categorized into tertiles (‘low’, ‘moderate’, and ‘high’). Age, sex, BMI, malignancy type, cumulative AC dose, and CTRCD risk score were evaluated as potential effect modifiers. Since all breast cancer patients in our cohort were female, malignancy type and sex were combined into a categorical variable (consisting of ‘males with a haematological malignancy’, ‘females with a haematological malignancy’, and ‘female breast cancer patients’), to isolate the potential effects of malignancy type and sex. Other malignancy types were excluded from this analysis given the low number of patients in this group. Age, BMI, and cumulative AC dose were added as continuous variables to test for effect modification and dichotomized to estimate easily interpretable hazard ratios (HRs). CTRCD risk score was determined using a modified version of the HFA-ICOS risk score and was collapsed into two categories (high/very high risk vs. low/moderate risk).²⁵ Baseline troponin and BNP/NT-proBNP levels were not included in the risk score calculation, since these were not routinely measured in the study period. Additionally, since the extent of smoking history was not quantified in most patients, any current or past smoking activity was attributed a risk point. Lastly, a sensitivity analysis excluding patients with acute myeloid leukaemia (AML) was performed, since this patient population typically presents with acute illness, for which chemotherapy is rapidly initiated at any hour of day, whereas treatments for most other malignancies are planned during office hours.

Statistical analysis

Associations between percentage of afternoon AC administration and time to CTRCD or HF were assessed using a parametric survival model based on Weibull distributions. This approach was specifically chosen as time to CTRCD and HF is interval censored. Additionally, these analyses were repeated using the grouped exposure variable (morning, intermediate, and afternoon group). Analyses were performed with and without corrections for potential confounders. Models without correction for confounders were used to construct survival curves. For CTRCD, the fully adjusted model included corrections for age, sex, malignancy type, cumulative AC dose, and mean infusion duration. For HF, only correction for cumulative AC dose was performed given the low number of events. The assumption for linearity was tested with restrictive cubic splines. To detect potential effect modification for the association between percentage of afternoon AC and CTRCD, interaction terms were added to both the model adjusted for cumulative dose and to the fully adjusted model. P-values for interaction terms were determined using the likelihood-ratio test, and interactions were deemed significant for P-values < 0.1. Survival analyses were performed using SAS v.9.4 (SAS Institute Inc.). All tables and forest plots were created in RStudio v.4.4.0 using the table1 and forestplot packages.

Results

Participants

A total of 270 subjects were included, consisting of 242 (89.6%) patients from the UMCU and 28 (10.4%) patients from the AZN. A flowchart of the participant selection process is depicted in Figure 1. Median age was similar amongst the study groups (Table 1). In 268 (99.3%) of included patients, all AC administration timepoints were available, in two patients one timepoint was missing. While baseline characteristics in the morning and afternoon groups were generally comparable, some differences were observed in the intermediate group. This group comprised the fewest patients (N = 32) and included a higher percentage of breast cancer patients, and, consequently, relatively more females and fewer patients with a haematological malignancy than the morning and afternoon groups. Supplementary material online, Table S1 lists all included malignancies in detail. Cumulative AC dose was similar in all groups (median 240 mg/m²). In the afternoon group, more patients had a mean AC infusion duration longer than 15 min (53.1%), compared with the morning (22.0%) and intermediate group (3.1%). Mean infusion durations ranged from 5 to 240 min. Supplementary material online, Figure S1 presents the distribution of percentage of ACs in the afternoon per study group. In total, 1361 AC cycles were administered. Doxorubicin was most administered (62.0%), followed by daunorubicin (19.9%), idarubicin (10.8%), mitoxantrone (3.9%) and epirubicin (3.4%).

Figure 1.

Flowchart of study population selection. AC, Anthracycline; AZN, St. Antonius Hospital Nieuwegein; UMCU, University Medical Center Utrecht.

Table 1.

Baseline characteristics

	Morning group	Intermediate group	Afternoon group
	(N = 91)	(N = 32)	(N = 147)
Females	43 (47.3%)	23 (71.9%)	75 (51.0%)
Age (years)a	56.0 [45.0, 69.0]	56.5 [45.3, 69.8]	55.0 [44.0, 64.0]
BMI (kg/m2)a	25.5 [22.3, 27.8]	24.8 [22.5, 27.8]	24.9 [22.6, 27.6]
Cardiovascular risk factors
Hypertension	23 (25.3%)	13 (40.6%)	32 (21.8%)
Hypercholesterolemia	16 (17.6%)	5 (15.6%)	19 (12.9%)
Missing	20 (22.0%)	5 (15.6%)	39 (26.5%)
Diabetes mellitus	6 (6.6%)	3 (9.4%)	12 (8.2%)
Smoking
Current smoker	10 (11.0%)	2 (6.3%)	20 (13.6%)
Former smoker	27 (29.7%)	8 (25.0%)	40 (27.2%)
Missing	13 (14.3%)	2 (6.3%)	11 (7.5%)
COPD	5 (5.5%)	2 (6.3%)	5 (3.4%)
Chronic kidney disease	6 (6.6%)	3 (9.4%)	6 (4.1%)
History of cardiac diseasea	18 (19.8%)	6 (18.8%)	30 (20.4%)
LV dysfunctiona	4 (4.4%)	1 (3.1%)	9 (6.1%)
HFA-ICOS risk score (modified)b
Low risk	35 (38.5%)	9 (28.1%)	73 (49.7%)
Moderate risk	25 (27.5%)	9 (28.1%)	33 (22.4%)
High risk	31 (34.1%)	13 (40.6%)	38 (25.9%)
Very high risk	0 (0%)	1 (3.1%)	3 (2.0%)
Socioeconomic statusc
Low	27 (29.7%)	9 (28.1%)	53 (36.1%)
Moderate	27 (29.7%)	19 (59.4%)	44 (29.9%)
High	35 (38.5%)	4 (12.5%)	49 (33.3%)
Missing	2 (2.2%)	0 (0%)	1 (0.7%)
Cardiovascular medication
Betablocker	8 (8.8%)	2 (6.3%)	24 (16.3%)
RAAS inhibitor	10 (11.0%)	6 (18.8%)	21 (14.3%)
Statin	12 (13.2%)	7 (21.9%)	21 (14.3%)
Oncological treatment characteristics
Malignancy
Haematological malignancy	64 (70.3%)	12 (37.5%)	102 (69.4%)
Breast cancer	19 (20.9%)	18 (56.3%)	28 (19.0%)
Other malignancy	8 (8.8%)	2 (6.3%)	16 (10.9%)
Haematological and other malignancy	0 (0%)	0 (0%)	1 (0.7%)
Number of AC cycles	5 [4, 6]	4 [4, 6]	5 [3, 6]
Cumulative dose of ACsd	240.0 [187.0, 300.0]	240.0 [205.0, 285.0]	240.0 [150.0, 270.0]
(mg/m²)
High cumulative dose of ACs	44 (48.4%)	8 (25.0%)	63 (42.9%)
(>240 mg/m2)
Prolonged AC infusion	20 (22.0%)	1 (3.1%)	78 (53.1%)
(>15 min)
Trastuzumab	8 (8.8%)	8 (25.0%)	17 (11.6%)
Left chest or mediastinal radiotherapy	10 (11.0%)	8 (25.0%)	15 (10.2%)
Missing	1 (1.1%)	0 (0%)	0 (0%)
Cumulative dose of left chest and mediastinal radiation (Gy)	43.2 [39.0, 54.0]	48.0 [33.0, 56.0]	42.6 [40.1, 64.0]
Missing	1 (5.3%)	0 (0%)	2 (5.4%)
Systemic corticosteroidse	82 (90.1%)	32 (100%)	132 (89.8%)
	0 (0%)	0 (0%)	3 (2.0%)
Melatonine
Benzodiazepinese	19 (20.9%)	5 (15.6%)	36 (24.5%)

Baseline characteristics of study participants, divided into three groups based on the percentage of cumulative anthracyclines received in the afternoon: morning group <50%, intermediate group 50%, and afternoon group > 50%. Categorical variables are presented as: number of patients (percentage of group). Continuous variables are presented as median [interquartile range]. Malignancies for which anthracycline treatment was given are presented. LV dysfunction was defined as a left ventricular ejection fraction < 50% or clinical heart failure with preserved ejection fraction.

AC, Anthracycline; Gy, Gray; COPD, Chronic obstructive pulmonary disease; HFA-ICOS, Heart Failure Association-International Cardio-oncology Society; LV, left ventricle; RAAS, Renin- Angiotensin-Aldosterone- System.

^aThese variables were determined before the first anthracycline treatment; all other variables were determined before the last anthracycline treatment.

^bHFA-ICOS risk score was calculated without data on troponin/(NT-pro)BNP levels, and any smoking history was attributed a risk point.

^cSocioeconomic status was derived by linking postal codes to published data from Statistics Netherlands.

^dConverted to doxorubicin equivalents.

^eDuring anthracycline treatment.

Outcomes

Median overall follow-up time was 3.9 years (interquartile range 2.0–6.3 years). In total, 66 (24.4%) of the 270 study participants developed CTRCD and 17 (6.3%) developed HF resulting from AC cardiotoxicity (Table 2). The incidence of CTRCD and HF was higher in the afternoon group than in the other two groups. Similarly, time to diagnosis was shortest in the afternoon group, for both CTRCD and HF. During the follow-up period, 98 patients died, and cause of death was mainly oncological (79.6%). In the morning, intermediate, and afternoon groups, the oncological mortality rates were 28.6%, 15.6%, and 32.0%, respectively. Descriptive statistics on causes of death can be found in Supplementary material online, Table S2. The association between administration time and mortality was not statistically tested given the oncological heterogeneity of the study population. Mortality numbers per malignancy type are provided in Supplementary material online, Table S3.

Table 2.

Descriptive statistics of CTRCD and heart failure

	Morning group	Intermediate group	Afternoon group
	(N = 91)	(N = 32)	(N = 147)
CTRCD
Events	14 (15.4%)	6 (18.8%)	46 (31.3%)
Follow-up time (years)	1.4 [0.7, 2.8]	1.2 [0.6, 2.7]	0.9 [0.3, 2.0]
Time to diagnosis (years)	0.7 [0.2, 2.4]	0.5 [0.3, 1.5]	0.4 [0.2, 0.8]
Heart failure
Events	3 (3.3%)	2 (6.3%)	12 (8.2%)
Follow-up time (years)	1.5 [0.9, 3.7]	1.8 [0.9, 3.2]	1.4 [0.6, 3.6]
Time to diagnosis (years)	1.4 [1.2, 4.3]a	1.2 [0.6, 1.9]a	0.5 [0.1, 6.4]a

Events are presented as: number of patients with event (percentage of group). Time periods are presented as median [interquartile range]. Only CTRCD and heart failure resulting from anthracycline cardiotoxicity are reported. Time to diagnosis was calculated using the date of the first anthracycline course as the starting point. Follow-up time represents either time to diagnosis or time to last follow-up, whichever occurred first. For CTRCD, the date of the last follow-up was the date of the last left ventricular ejection fraction (LVEF) measurement (by any modality), while for heart failure this consisted of either the date of last LVEF measurement or the last clinical visit.

CTRCD, cancer therapy-related cardiac dysfunction.

^aInterquartile range could not be determined; therefore, minimum and maximum values are reported.

Compared with the morning group, patients in the afternoon group had a higher risk of developing CTRCD adjusted HR 2.88 (95% CI: 1.52–5.44) (Table 3). Additionally, an association was found when including percentage of afternoon AC administration as a continuous variable: adjusted HR 1.14 (95% CI: 1.04–1.24) for every additional 10% of ACs administered in the afternoon. Additional adjustment for SES resulted in similar HRs (see Supplementary material online, Table S4). Similarly, additional adjustment for trastuzumab did not affect the HRs. No effect modification (all P-values > 0.1) was observed for any of the tested variables (Figure 2).

Table 3.

Hazard ratios for developing anthracycline-induced CTRCD and heart failure

	Per 10% AC in afternoon	Afternoon vs. morning	Intermediate vs. morning
CTRCD
Events	66/270	46/147 vs. 14/91	6/32 vs. 14/91
Adjusted HRa (95% CI)	1.14 (1.04–1.24)	2.88 (1.52–5.44)	1.86 (0.69–4.99)
Heart failure
Events	17/270	12/147 vs. 3/91	2/32 vs. 3/91
Adjusted HRb (95% CI)	1.19 (1.01–1.41)	2.64 (0.74–9.40)	1.93 (0.32–11.63)

Events are presented as: number of patients with event (percentage of group).

CTRCD, cancer therapy-related cardiac dysfunction; HR, hazard ratio.

^aModel was adjusted for age, biological sex, malignancy type (haematological cancer, breast cancer, other malignancy), cumulative anthracycline dose, and mean infusion duration.

^bModel was adjusted for cumulative anthracycline dose.

Figure 2.

Forest plot of sensitivity analyses. Hazard ratios with 95% confidence intervals for developing CTRCD per 10% increase of anthracyclines administered in the afternoon. Hazard ratios are presented for all patients and for subgroups based on age, body mass index, biological sex, malignancy type, and adjusted HFA-ICOS risk score. HFA-ICOS risk score was adjusted by calculating the score without data on troponin and BNP/NT-proBNP levels, and by assigning a risk point for any smoking history. 95% CI, 95% confidence interval; AC, Anthracycline; BMI, Body mass index; HFA-ICOS, Heart Failure Association—International Cardio-oncology Society; HR, hazard ratio; CTRCD/N, Number of patients with CTRCD per total number of patients in each group. *The P-value represents the P-value for the interaction term added to the fully adjusted model. For age and BMI, the interaction term was added as a continuous variable.

Afternoon AC administration was also associated with a higher risk of developing HF: the adjusted HR for developing HF was 1.19 (95% CI: 1.01–1.41) for every 10% of ACs administered in the afternoon (Table 3). In the grouped analyses, the adjusted HR for developing HF in the afternoon group was not statistically significant: 2.64 (95% CI: 0.74–9.40) compared with the morning group. Since event count for HF was low, potential effect modification was not assessed. Graphical presentations of time to CTRCD and HF are shown in Figure 3.

Figure 3.

Survival plots. Survival plots for (A) CTRCD free survival (B) HF free survival in the three anthracycline administration time groups. Morning group (N = 91) > 50% of anthracyclines administered in the morning (12 a.m.–11:59 a.m.) (Solid line). Afternoon group (N = 147): > 50% of anthracyclines administered in the afternoon (12 p.m.–11:59 p.m.) (Evenly dashed line). Intermediate group (N = 32): 50% of anthracyclines administered in the afternoon and 50% in the morning (Unevenly dashed line). Of note, due to the complexity of our statistical model, these curves were constructed without correction for competing risks (i.e. mortality), and are therefore not numerically interpretable, however they provide a visual representation of the impact of AC administration timing on CTRCD free survival and HF free survival. CTRCD, Cancer therapy-related cardiac dysfunction; HF, Heart failure.

Sensitivity analyses excluding patients with AML showed comparable results for the association between afternoon AC administration and risk of developing CTRCD and HF (see Supplementary material online, Table S5).

Discussion

This observational cohort study found that afternoon AC administration is strongly associated with a higher risk of developing CTRCD and HF. Furthermore, the association with CTRCD was unaffected by age, sex, BMI, malignancy type, cumulative dose, and CRS stratification.

To the best of our knowledge, this is the first clinical study with the primary focus on investigating the administration-time-dependency of AC cardiotoxicity. Previous RCTs studying different classes of chemotherapeutics, have shown that chronomodulated chemotherapy has the potential to decrease toxicity in general, while maintaining oncological efficacy.⁷ Only one RCT investigated the time-of-day effect of AC + cisplatin administration on anti-cancer efficacy and reported on cardiotoxicity as a safety outcome. This study, involving patients with endometrial cancer, found no statistically significant differences in cardiotoxicity or efficacy between the chronomodulated intervention arm (doxorubicin at 6 a.m. + cisplatin at 6 p.m.) and the control arm (doxorubicin, immediately followed by cisplatin, at random times during office hours).²⁶ The absence of an observed effect on cardiotoxicity may be due to a higher cumulative dose of doxorubicin in the intervention arm (median 246 mg/m² vs. 209 mg/m²). Additionally, methodological limitations were present hindering adequate isolation of the effect of the time-of-day.²⁷ Moreover, since only symptomatic HF was considered, the number of events was low, possibly hindering detection of an effect.

Our findings suggest circadian variation in cardiac sensitivity to AC-induced damage. This corroborates earlier preclinical studies, in which a connection between timing of AC administration and cardiotoxicity was observed both in vitro and in vivo.^8,9,10 A possible explanation was provided by Yang et al.¹⁰ who found that AC-induced cardiotoxicity in rodents is associated to the 24-h variations in the activity levels Sirtuin 3, through regulation of mitochondrial function and ROS production. Although this may partly explain the circadian variation in cardiac sensitivity to AC-induced damage, it is likely that additional mechanisms are involved. To reach full understanding, the potential 24-h variation of other AC-cardiotoxicity mechanisms needs to be explored. In addition, day-night fluctuations in AC pharmacokinetics leading to variations in serum concentrations, have been reported.²⁸ Larger studies are needed to determine the clinical significance of these fluctuations and their potential contribution to an administration-time-dependent effect.

Maintaining or improving oncological efficacy remains the primary goal of chemotherapeutic treatment. Since our cardio-oncology outpatient population is heterogeneous regarding malignancy type and stage, we did not assess the association between administration time and oncological endpoints and other toxicities. Although it is not possible to draw definitive conclusions regarding oncological efficacy, our survival data suggest that oncological mortality was similar in the morning and afternoon group. The lowest mortality rate was observed in the intermediate group, which may be explained by the relatively high proportion of breast cancer patients in this group, who generally have a better prognosis than patients with a haematological malignancy.²⁹ Our findings are in line with results of previous RCTs showing that efficacy was not decreased by adjusting the administration time of chemotherapeutics.⁷ On the other hand, a recent observational study found that morning administration of an AC-containing chemotherapy regimen for diffuse large B cell lymphoma was associated with lower overall survival in females, whereas this association was not present in males.³⁰ Similarly, a meta-analysis comparing chronomodulated, non-AC containing, chemotherapy to conventional chemotherapy for metastatic colorectal cancer, found that males had a longer overall survival when receiving chronomodulated chemotherapy, while in females this led to worse survival outcomes.³¹ In our cohort, no sex-differences were observed regarding oncological survival (data not shown). However, our study was not designed for detailed quantification of oncological outcomes. Therefore, future studies are needed to further explore the impact of AC administration time on oncological efficacy and other toxicities, holding sufficient power to detect potential sex-differences in these outcomes.

Limitations

The observational nature of this study inevitably leads to a risk of bias. Even though this was minimized by adjusting for potential confounders, the presence of unmeasured confounders could not be excluded. Another limitation lies in the generalizability of our findings; since commonly not every patient receiving ACs is referred to a cardio-oncology outpatient clinic. Hence, our study population comprised not all patients treated with AC at an oncology department but a selected population, which might probably be at a higher risk of cardiotoxicity.¹⁸ However, the sensitivity analysis on HFA-ICOS risk score suggests that the association between administration time and CTRCD is not dependent on patients’ risk profiles, since similar HRs were found for groups of lower and higher risk scores.

Clinical implications

This study shows that morning administration of ACs may be an effective measure in preventing AC-induced cardiotoxicity. Since cancer incidence is rising, and is expected to keep rising, our findings are relevant for a growing population.⁶ Current pharmacological interventions to decrease cardiotoxicity can introduce additional toxicities, such as myelotoxicity in the case of dexrazoxane, indicating that a non-pharmacological intervention may prove to be more favourable. In addition, it is important to consider innovative strategies to regulate the increasing health costs. Whereas pharmacological interventions for preventing CTRCD contribute to the growing economic strain of oncological care, chronomodulated chemotherapy comes at no additional costs, making it a highly accessible measure for any medical centre. Nevertheless, implementation would require significant logistical adjustments.

Conclusions

Our findings indicate that morning administration of ACs is associated with a lower risk of developing AC-induced CTRCD and HF in patients with cancer. Furthermore, for CTRCD, this association was consistent across the sensitivity analyses of cardiotoxicity predictors. Given that ACs are currently administered at not stipulated times of day, adjusting their timing to the morning appears to be a safe and effective approach to minimize cardiac damage, at no additional costs. However, before clinical implementation can take place, replication of our findings is warranted, preferably in the form of an RCT.

Lead author biography

Dr Markella I. Printezi is a medical doctor with clinical and research interests in cardiology, oncology and circadian biology. Her work focuses on the role of biological clocks in heart failure and the timing of cardiotoxic cancer therapies.

Data availability

Data supporting this study cannot be made available, since the included patients did not provide consent for publication of their data. Fully anonymized aggregated data can be made available upon request.

Supplementary material

Supplementary material is available at European Heart Journal Open online.

Author contributions

Markella I. Printezi (Conceptualization [supporting]; Data curation [lead]; Formal analysis [supporting]; Investigation [lead]; Methodology [equal]; Project administration [lead]; Validation [lead]; Visualization [lead]; Writing—original draft [lead]), Arco J. Teske (Data curation [supporting]; Methodology [equal]; Supervision [equal]; Validation [equal]; Writing—review & editing [lead]), Nicolaas P.A. Zuithoff (Formal analysis [lead]; Validation [equal]; Visualization [equal]; Writing—review & editing [equal]), Kim Urgel (Data curation [supporting]; Investigation [supporting]; Validation [equal]; Writing—review & editing [equal]), Rhodé M. Bijlsma (Validation [equal]; Writing—review & editing [equal]), Anna van Rhenen (Validation [equal]; Writing—review & editing [equal]), Maarten Jan Cramer (Validation [equal]; Writing—review & editing [equal]), Cornelis J.A. Punt (Validation [equal]; Writing—review & editing [equal]), Anne M. May (Conceptualization [equal]; Methodology [equal]; Supervision [lead]; Validation [lead]; Writing—review & editing [lead]), and Linda W. van Laake (Conceptualization [lead]; Funding acquisition [lead]; Methodology [equal]; Supervision [lead]; Validation [lead]; Writing—review & editing [lead])

Funding

This work was supported by Hartstichting: Dekker Senior Clinical Scientist 2019, grant number 2019T056.