Redefining the Incidence and Profile of Fluoropyrimidine-Associated Cardiotoxicity in Cancer Patients: A Systematic Review and Meta-Analysis

Aim: The cardiac toxicity that occurs during administration of anti-tumor agents has attracted increasing concern. Fluoropyrimidines have been used for more than half a century, but their cardiotoxicity has not been well clarified. In this study, we aimed to assess the incidence and profile of fluoropyrimidine-associated cardiotoxicity (FAC) comprehensively based on literature data. Methods: A systematic literature search was performed using PubMed, Embase, Medline, Web of Science, and Cochrane library databases and clinical trials on studies investigating FAC. The main outcome was a pooled incidence of FAC, and the secondary outcome was specific treatment-related cardiac AEs. Random or fixed effects modeling was used for pooled meta-analyses according to the heterogeneity assessment. PROSPERO registration number: (CRD42021282155). Results: A total of 211 studies involving 63,186 patients were included, covering 31 countries or regions in the world. The pooled incidence of FAC, by meta-analytic, was 5.04% for all grades and 1.5% for grade 3 or higher. A total of 0.29% of patients died due to severe cardiotoxicities. More than 38 cardiac AEs were identified, with cardiac ischemia (2.24%) and arrhythmia (1.85%) being the most frequent. We further performed the subgroup analyses and meta-regression to explore the source of heterogeneity, and compare the cardiotoxicity among different study-level characteristics, finding that the incidence of FAC varied significantly among different publication decades, country/regions, and genders. Patients with esophagus cancer had the highest risk of FAC (10.53%), while breast cancer patients had the lowest (3.66%). The treatment attribute, regimen, and dosage were significantly related to FAC. When compared with chemotherapeutic drugs or targeted agents, such a risk was remarkably increased (χ2 = 10.15, p < 0.01; χ2 = 10.77, p < 0.01). The continuous 5-FU infusion for 3–5 consecutive days with a high dosage produced the highest FAC incidence (7.3%) compared with other low-dose administration patterns. Conclusions: Our study provides comprehensive global data on the incidence and profile of FAC. Different cancer types and treatment appear to have varying cardiotoxicities. Combination therapy, high cumulative dose, addition of anthracyclines, and pre-existing heart disease potentially increase the risk of FAC.


Introduction
Fluoropyrimidines including 5-FU, its oral pro-drug capecitabine, the recent compound preparations S-1 (tegafur), and TAS 102 have been used in the treatment of solid tumors for more than half a century [1,2]. Although numerous therapeutic strategies have been introduced in recent years, such as targeted therapy [3] and immunotherapy [4], fluoropyrimidines are still the fundamental chemotherapeutic agents for treating many The main outcome was the incidence of the cardiotoxicities for all grades, grade 3 or higher, and grades 1-2. The profiles of cardiac AEs were also prespecified as important secondary outcomes. The data of basic characteristics, treatment details, and clinical results of cardiotoxicity were obtained from each included study.

Statistical Analysis
The meta-analyses were conducted using R software (Version 4.0.6) with "meta", "rmeta", and "metafor" packages. Shapiro-Wilk normality tests on the raw rate and transformed data (log, logit, arcsine, and Freeman-Tukey transformation) were used to determine the most appropriate data type for the pooled analysis. The inter-study heterogeneity was detected by the Cochran's Q test reporting I 2 statistic and p values. A random-effect model was adopted in case of an indication of significant heterogeneity (I 2 > 50% or p < 0.1) [19,20], otherwise, the fixed-effect model was used. Subgroup analyses were performed based on study-level moderators in order to compare the incidence of FAC among studies with different characteristics. A multilevel meta-regression analysis was conducted to detect the source of heterogeneity further and examine the influence of the moderator variables. Funnel plot and Egger's tests were used to assess publication bias. The sensitivity analysis was performed by excluding studies one by one to determine the stability of the results of the meta-analysis.

Eligible Studies and Characteristics
A total of 211 eligible studies involving 63,186 patients were included [15, ( Figure 1). Table 1 summarizes the characteristics of the included studies. The involving population covered 31 countries or regions in the world ( Figure 2). According to the NIH quality assessment tools, 61 articles (28.9%) had a good quality score, 150 (71.1%) fair quality, and none was classified as poor (high risk of bias). The detailed information of each included study is shown in Supplementary Table S3.

Basic Demographic and Study-Level Factors
The results of the subgroup analysis are shown in Figure 4. The incidence of FAC has significantly increased in the past three decades (χ 2 = 7.8, p = 0.02). Studies conducted in Asia outlined a higher incidence than in Europe (χ 2 = 4.44, p = 0.03) and America (χ 2 = 4.45, p = 0.03). No significant difference was observed between their subgroups in terms of study design, trial phase, population age, and methodological quality. Sixty-three studies only included females with breast cancer, producing a lower incidence of FAC than those involving both females and males (χ 2 = 8.75, p < 0.01). Studies that excluded people with pre-existing cardiac disorders had a reduced incidence of FAC (χ 2 = 4.29, p = 0.04).

Basic Demographic and Study-Level Factors
The results of the subgroup analysis are shown in Figure 4. The incidence of FAC has significantly increased in the past three decades (χ 2 = 7.8, p = 0.02). Studies conducted in Asia outlined a higher incidence than in Europe (χ 2 = 4.44, p = 0.03) and America (χ 2 = 4.45, p = 0.03). No significant difference was observed between their subgroups in terms of study design, trial phase, population age, and methodological quality. Sixty-three studies only included females with breast cancer, producing a lower incidence of FAC than those involving both females and males (χ 2 = 8.75, p < 0.01). Studies that excluded people with pre-existing cardiac disorders had a reduced incidence of FAC (χ 2 = 4.29, p = 0.04).

FAC for Different Treatment Parameters
Significant differences in FAC incidence existed between treatment parameters. Fluoropyrimidines for advanced/metastatic/relapsed diseases had a higher incidence of FAC compared with neoadjuvant or adjuvant treatments (χ 2 = 6.91, p = 0.03); however, no statistical difference was observed between different treatment lines (first-line vs. ≥ second-line, χ 2 = 0.05, p = 0.82). The 5-FU induced cardiotoxicity in monotherapy was significantly lower than that in the combination therapy, either combined with other chemotherapeutic drugs (χ 2 = 10.15, p < 0.01) or targeted agents (χ 2 = 10.77, p < 0.01).

The Results of Meta-Regression Analysis
A meta-regression analysis was performed to identify the factors influencing FAC incidence further. The univariable meta-regression of continuous data revealed that female proportion (negative, Q = 8.59, p < 0.01) and 5-FU dosage (positive, Q = 9.57, p < 0.01) were strongly correlated with FAC (Supplementary Figure S7A,B). Publication year and median age of population were also correlated with the cardiotoxicity, but the differences were not significant (p = 0.058; p = 0.071) (Supplementary Figure S7C,D). Results of the multilevel meta-regression analysis are shown in Table 3. As can be seen in the results, the moderators of publication year (p = 0.02), country/region (p < 0.01), pre-existing cardiac disorders (p < 0.01), study design (p = 0.024), treatment attribute (p < 0.001), cancer type (p = 0.026), regimen (p = 0.043), and anthracycline combination (p = 0.027) were significant predictors influencing the occurrence of cardiotoxicities. This multilevel meta-regression model totally explained more than half of the inter-study heterogeneity (R 2 = 51.74%).

Publication Bias and Sensitivity Analysis
No obvious asymmetry was observed in the funnel plots of the main outcomes, suggesting no evidence of significant publication bias, which was confirmed by the Egger's test (Supplementary Figure S8A-F). The results of the sensitivity analysis showed that no individual study substantially influenced the pooled results of the above main outcomes (Supplementary Figure S9A-F), indicating that the results of this meta-analysis were relatively stable.

Discussion
This study generated robust epidemiological data of FAC incidence and profile based on a single-rate meta-analysis of 211 studies and 63,186 patients, which revised the previous over-or under-estimation of FAC [31,38,231]. We further identified several factors influencing the occurrence of FAC through subgroup analysis and meta-regression analysis. Cardiotoxicity caused by fluoropyrimidines has not been widely recognized and studied in the past. However, along with the wide application of fluoropyrimidine-combined therapy and the increasing demand of long-term mediation for advanced-stage patients with improved survival, increasingly attention has been paid to the cardiotoxicity problem of fluoropyrimidines. The incidence of FAC (5.04%), by our meta-analytic, is second only or even comparable to the familiar cardiotoxicity of anthracyclines (data from a meta-analysis based on 50 thousand cases: The incidence of 6.3%) [232]. Indeed, the real prevalence of FAC might be even higher than what we reported, since not all included studies undertook the most comprehensive cardiac evaluation, which might have resulted in missed FAC detections. Therefore, FAC must be given a high level of attention in clinical practice, as capecitabine, 5-FU, and its modified agents are increasingly used in maintenance therapy for malignancies. The management of FAC in cancer patients has a tremendous impact on the type of antitumor therapies, as well as long-term morbidity and mortality [233].
The profile of fluoropyrimidine-related cardiac AEs has been clearly depicted in our results, which provided evidence for close monitoring and early identification of pertinent cardiac symptoms and signs. Cardiac ischemia is the typical and most common FAC, occurring in 2.24% of patients, and one of the most common causes of fluoropyrimidine-related death (27.19%). The typical ST-T ECG changes were observed in most patients with cardiac ischemia. However, the overall incidence of ECG alterations was even higher than the incidence of cardiac AEs (5.85% vs. 5.04%), indicating that some subjects presented asymptomatic ECG changes, which is often a warning signaling an imminent ischemic event. Several studies considered patients who underwent 24 h ECG Holter monitoring, and, as a result, ischemic ECG changes were observed in 31-68% of patients [233,234]. Therefore, continuous ECG monitoring should be strongly recommended in patients subjected to fluoropyrimidines as the simplest and most useful method for early recognition of FAC. Arrhythmia, including atrial fibrillation, tachycardia, bradycardia, and conduction disorder were also indicated as the common features of FACs (1.85%), which might be a result of cardiac apoptosis triggered by ischemic change and cardiomyocyte damage [235]. LVEF can be used as an indicator of cardiac pump function, which is closely related to heart failure [236]. Our results showed a reduction in LVEF ≥ 20% in 1.5% of patients, while 0.91% developed cardiac dysfunction. However, the evaluation of LVEF using echocardiography is limited by the intrinsic operator-dependency of this technique. Thus, multi-level detection (e.g., ECG, echocardiography, myocardial enzyme, fMRI, coronary angiography, and radionuclide ventriculography) would be helpful for improving the sensitivity required to identify FAC.
Pre-treatment FAC risk assessment should ideally be performed, and the HFA-ICOS risk assessment tools can be considered [233]. Several studies suggested that FACs varied with the population characteristics, cancer types, and treatment regimens [237]. In contrast, other studies found that these factors were insignificant [95]. Our study performed a comprehensive sub-analysis to assess the influences of these variations on FAC, which is useful for indicating the risk factors of FAC. Being female was reported to be a protective factor for myocardial infarction and ischemia [238]. As shown in our results, the incidence of FAC decreased as the proportion of females increased, and the female-only population had a significant lower incidence than the general population. The lower cardiotoxicity was probably due, to some extent, to the protective effect of female hormones. On the other hand, it could also be related to the specific cancer type and regimens, since all patients in the female-only subgroup suffered from breast cancer, and most of them underwent capecitabine-based regimens (1000 mg/m 2 , d1-14, Q3w).
Previous studies reported a higher incidence of FAC in the elderly population [42], and it was reported that patients with an age ≥ 80 years are at great risk of treatment-related cardiac problems [233]. Similarly, we found a positive correlation between median age and cardiotoxicity, but no significant difference was detected in the subgroup analysis. Moreover, only four studies were included in the elderly subgroup (≥60 years) involving only 260 patients. Thus, it was not possible in this work to demonstrate whether the elderly is at particularly risk of FAC. Further large-scale stratified analyses focusing on the elderly are required. Notably, studies performed in Asia showed a higher incidence of cardiotoxicities than those in Europe and America. This result was consistent with the previous findings of Peng et al. [44], in which the incidence of cardiotoxicity in the Chinese population was higher than that in the non-Chinese population (25% vs. 19.9%). Such discrepancies may be derived from the genetic polymorphisms characterizing different ethnicities. For example, the frequency of DPD enzymatic activity varies greatly among Asian, Caucasian, and African-American populations, and it is an indicator of a remarkable risk for cardiac damage [239]. In addition, it is not surprising that pre-existing cardiac disorder significantly increased the risk of cardiotoxicities. One potential reason is that the unrecovered disease causes damage to the cardiomyocytes, making them more sensitive to an external stimulus like 5-FU infusion. Therefore, special care should be taken when administering fluoropyrimidines to patients with underlying cardiac disorders.
Our results indicated that the incidence of FAC differed among tumors. The highest incidence of cardiotoxicities was observed in esophageal cancer (10.53%), whereas the lowest was found in breast cancer (3.66%). Meydan's study also showed a similar result [124], in which no patient with breast cancer suffered from cardiotoxicity, while the incidence in other tumors, such as colorectal cancer, was more than 3.5%. The discrepancy in cardiotoxicity among tumors may also be derived from gender, treatment regimen, or dosage schedule. For example, the addition of radiotherapy for esophageal cancer greatly increases the treatment-related cardiotoxicity. However, further studies on the FAC among different tumors are required because of the lack of direct comparisons between tumors.
The occurrence of cardiotoxicity varied among different fluoropyrimidine regimens. Combination therapy with other chemotherapy agents or targeted drugs resulted in a higher incidence of FAC than monotherapy. The increased cardiotoxicity could be due to the additional or synergistic deleterious cardiac effects from multiple agents. For example, it is worth noting that the cardiotoxicity significantly increased when fluoropyrimidine was coupled with cardiotoxic drugs such as anthracyclines, which has been approved as a risk for the occurrence of adverse cardiac events [233]. Therefore, when chemotherapy regimens containing anthracyclines are considered, close attention should be paid. In addition, our findings observed that cardiotoxicity induced by capecitabine was not significantly higher than 5-FU, although some previous studies showed a higher toxicity in capecitabine treatment [44,95]. The incidence of FAC differs by 5-FU administration (continuous infusion or bolus infusion), and it occurs in a cumulative dose-dependent manner [147], which was confirmed by our results. In view of the prominent cardiotoxicity of fluoropyrimidines, it is important to develop safe and efficacious new fluoropyrimidine drugs. TAS-102, a novel fluorouracil agent, is a promising safe option for patients. In addition, raltitrexed (Tomudex ® ) can also be used as an alternative to conventional fluorouracil drugs, with a lower incidence of FAC [240].
This study contains some limitations. The first is the heterogeneity at the study level. Although series subgroup analyses and meta-regression were performed to explore the source of heterogeneity further, they did not explain all the heterogeneity of the pooled effects. The clinical heterogeneity of the participants in the included studies could potentially have induced heterogeneity in the results of this meta-analysis. The attributes of the included studies could also have led to unstable results. For example, prospective RCT research can produce more robust conclusions in theory. Secondly, although all the reported cardiac AEs were diagnosed and graded based on the general NCI or WHO criteria, cardiotoxicity might have been underestimated in some studies due to the lack of cardiac-specific examinations. Finally, most of the included studies, especially the clinical trials, adopted strict inclusion criteria, limiting the generalizability of our results for patients outside the inclusion criteria (e.g., patients younger than 18 years or older than 70 years). Hence, real-word studies based on a large-scale population and robust analysis should be necessary to confirm our findings.

Conclusions
In conclusion, this meta-analysis systematically and comprehensively redefined the incidence and profile of FAC. FAC is not as rare as it seems, and involves a wide variety of related cardiac AEs, with cardiac ischemia and arrhythmia being the most common. The occurrence of cardiotoxicity varies among different publication decades, country/regions, genders, cancer types, and treatment details. Combination therapy, high cumulative dose, addition of anthracyclines, and pre-existing heart disease potentially increase the risk of FAC. This global overview of cardiotoxicity in fluoropyrimidine-based treatment may be used as a clinical reference in clinical practice for the management of cancer therapy.   Figure S7: The bubble plots showed the results of univari-able meta-regression of continuous data (female proportion, 5-FU dosage schedule, publication year, and median age). A. significant negative relationship between female proportion and the incidence of cardiotoxicities (Q = 8.59, p < 0.01); B. significant positive relationship between 5-FU dosage schedule and the incidence of cardiotoxicities (Q = 9.57, p < 0.01); C. the relationship between publication year and the incidence of cardiotoxicities (Q = 3.69, p = 0.058); D. the relationship between median age and the incidence of cardiotoxicities (Q = 3.33, p = 0.071); Supplementary Figure S8: Egger's funnel plot for publication bias test of the main outcomes, and no evidence of significant publication bias was detected. A. incidence of all-grade FACs; B. incidence of grade 3 or higher FACs; C. incidence of grade 1-2 FACs; D. incidence of cardiac ischemia; E. incidence of arrhythmia; F. incidence of ECG changes; Supplementary Figure S9: Forest plot for the sensitivity analysis of main outcomes. Removal of individual studies one at a time did not materially alter the results. A. incidence of all-grade FACs; B. incidence of grade 3 or higher FACs; C. incidence of grade 1-2 FACs; D. incidence of cardiac ischemia; E. incidence of arrhythmia; F. incidence of ECG changes.