How Asian Breast Cancer Patients Experience Unequal Incidence of Chemotherapy Side Effects: A Look at Ethnic Disparities in Febrile Neutropenia Rates

Simple Summary In this two-part study, we assessed the risk factors associated with FN. On top of commonly associated risk factors such as age and body mass index, we found that there are ethnic differences in terms of FN incidence. Notably, patients of non-Chinese ethnicity had a higher FN incidence. This trend remained significant even after multiple adjustments with other patients, tumors, and treatment characteristics. Further analysis exploring differences across ethnicities revealed that Indian patients were more likely to experience greater dips in absolute neutrophil count during chemotherapy and Malay patients were more likely to be administered with anthracyclines, both of which contribute to higher FN risk. Importantly, we also found that non-Chinese patients were more likely to experience multiple FN episodes during chemotherapy. Additional research is needed to investigate the possibility of pharmacogenetic differences across these ethnicities. Abstract The majority of published findings on chemotherapy-induced febrile neutropenia (FN) are restricted to three ethnic groups: Asians, Caucasians, and African Americans. In this two-part study, we examined FN incidence and risk factors in Chinese, Malay, and Indian chemotherapy-treated breast cancer (BC) patients. Hospital records or ICD codes were used to identify patients with FN. In both the Singapore Breast Cancer Cohort (SGBCC) and the Joint Breast Cancer Registry (JBCR), the time of the first FN from the start of chemotherapy was estimated using Cox regression. Multinomial regression was used to evaluate differences in various characteristics across ethnicities. FN was observed in 170 of 1014 patients in SGBCC. The Cox model showed that non-Chinese were at higher risk of developing FN (HRMalay [95% CI]:2.04 [1.44–2.88], p < 0.001; HRIndian:1.88 [1.11–3.18], p = 0.018). In JBCR, FN was observed in 965 of 7449 patients. Univariable Cox models identified ethnicity, a lower baseline absolute neutrophil count, non-luminal A proxy subtypes, and anthracycline-containing regimens as risk factors. Disparities across ethnicities’ risk (HRMalay:1.29 [1.07–1.54], p = 0.006; HRIndian:1.50 [1.19–1.88], p < 0.001) remained significant even after further adjustments. Finally, an age-adjusted multinomial model showed that Malays (p = 0.006) and Indians (p = 0.009) were significantly more likely to develop multiple episodes of FN during treatment. Ethnic differences in chemotherapy-induced FN among BC patients exist. Further studies can focus on investigating pharmacogenetic differences across ethnicities.


Introduction
Febrile neutropenia (FN) is a prominent hematological adverse effect of chemotherapy that leads to dosage decrease and treatment delay [1]. Both are associated with a decrease in relative dose intensity, which hinders the recovery of early-stage breast cancer patients [1][2][3]. Furthermore, FN is a potentially fatal toxicity, akin to sepsis, and may necessitate emergency admission and hospitalization [4,5].
The occurrence of chemotherapy-induced FN may be associated with ethnic origin, with Asians being more vulnerable to FN than patients of European ancestry [6][7][8][9][10][11][12]. A review done by O'Donnell found that East Asians are more susceptible to chemotherapy side effects, such as neutropenia [13]. In conjunction, a meta-analysis exploring the incidence of docetaxel-induced severe neutropenia between Asian and non-Asian clinical studies found notable differences between the two populations [6]. Another multi-center study reported that racial differences in acute toxicity exist in breast cancer patients treated with fluorouracil, epirubicin, and cyclophosphamide (FEC) 100 chemotherapy [8]. It should be highlighted that the majority of published findings on cancer pharmaco-ethnicity are restricted to three ethnic groups: Asians, Caucasians, and African Americans [13]. Hence, there is a need to further explore the prevalence and risk factors of FN in other ethnic groups or geographic settings [7,11,14].
In this two-part study, we first examined the incidence of chemotherapy-induced FN in a retrospective cohort study comprising 1014 breast cancer patients in Singapore. Using the independent Joint Breast Cancer Registry, we further studied the incidence of FN amongst 7449 chemotherapy-treated patients as well as associations between ethnicity and other patients, tumor, and treatment characteristics.

Materials and Methods
This study comprises two independent datasets from three health clusters in Singapore: National University Health System (NUHS), National Healthcare Group (NHG), and Singapore Health Services (SingHealth). We explored the associations between ethnicity and FN occurrence in breast cancer patients recruited from NUHS and NHG, using breast cancer patients recruited from the Singapore Breast Cancer Cohort (SGBCC) (n = 1014). To reach a deeper understanding of our findings, we studied various potential confounders in a larger registry dataset from the SingHealth cluster, the Joint Breast Cancer Registry (JBCR) (n = 7449).

Singapore Breast Cancer Cohort (SGBCC) Dataset
Briefly, SGBCC is a multi-institutional breast cancer cohort established in 2009. The seven recruiting public hospitals, namely, National University Hospital (NUH), KK Women's and Children's Hospital (KKH), Tan Tock Seng Hospital (TTSH), National Cancer Centre Singapore (NCCS), Singapore General Hospital (SGH), Changi General Hospital (CGH), and Ng Teng Fong General Hospital (NTFGH), collectively treat~76% of all breast cancer patients in Singapore; the details on the cohort profile can be found in this paper [15]. Patients from NUH and TTSH were included in the analytical cohort. FN cases were identified via hospital admission date where the laboratory test has a neutrophil count of fewer than 500 cells/µL together with a single oral temperature greater than or equal to 38.3 • C or a temperature greater than or equal to 38 • C for at least an hour during or within 30 days from the last chemotherapy session. Univariable Cox regression models were used to assess the associations between the various risk factors and FN. Multivariable

Joint Breast Cancer Registry (JBCR)
The Joint Breast Cancer Registry (JBCR) is a hospital-based cancer registry that holds a wealth of information on breast cancer patients diagnosed or treated at hospitals within the SingHealth Cluster in Singapore [16]. The sites include SGH, NCCS, KKH, CGH, and Seng Kang General Hospital (SKH). Data collected by JBCR include patient demographics, primary tumor information (e.g., diagnosis date, tumor histology, stage at diagnosis, tumor size, nodal involvement, hormonal status), treatment details, adverse drug reactions observed, survival status, and treatment-related costs.
JBCR was approved by the SingHealth Centralized Institutional Review Board (CIRB Ref: CIRB2019/2419 and 2016/3010). As the study relies on deidentified registry-based data, the need to obtain informed consent was waived. This study was approved by the A*STAR Institutional Review Board (2020-005).

Outcomes of Interest
In JBCR, patients with FN were identified either via hospital admission date or Accident and Emergency visit date with confirmation of ICD FN diagnosis (fever: ICD9/10:2880/D70 and neutropenia: ICD9/10:7806/R509). The primary outcome of interest was the incidence of FN during or within 30 days from the last chemotherapy session. The secondary outcome of interest was the total number of FN occurrences during the same time period. Details on the definition of study time points can be found in Supplementary Figure S2.

Patient Characteristics
Patient characteristics that were included in this analysis were treatment institution, age at diagnosis, year of diagnosis, ethnicity (Chinese/Malay/Indian), baseline and changes in body mass index (BMI), and absolute neutrophil count (ANC). Information on time points of BMI and ANC records used is given in Supplementary Figure S2. It should be noted that a large proportion of participants had incomplete records for baseline BMI, changes in BMI, baseline ANC, and changes in ANC.

Outcomes of Interest
In JBCR, patients with FN were identified either via hospital admission date or Accident and Emergency visit date with confirmation of ICD FN diagnosis (fever: ICD9/10:2880/D70 and neutropenia: ICD9/10:7806/R509). The primary outcome of interest was the incidence of FN during or within 30 days from the last chemotherapy session. The secondary outcome of interest was the total number of FN occurrences during the same time period. Details on the definition of study time points can be found in Supplementary Figure S2.

Patient Characteristics
Patient characteristics that were included in this analysis were treatment institution, age at diagnosis, year of diagnosis, ethnicity (Chinese/Malay/Indian), baseline and changes in body mass index (BMI), and absolute neutrophil count (ANC). Information on time points of BMI and ANC records used is given in Supplementary Figure S2. It should be noted that a large proportion of participants had incomplete records for baseline BMI, changes in BMI, baseline ANC, and changes in ANC.

Treatment Details
Treatment characteristics included surgery type (mastectomy/breast conserving/ unknown/no surgery/missing), radiotherapy (yes/no/missing), tamoxifen treatment (yes/no/missing), herceptin treatment (yes/no/missing), chemotherapy type (neoadjuvant/adjuvant), and granulocyte colony-stimulating factor (GCSF) use (yes/no). Individual GCSF time points (before/during chemotherapy) were derived based on the recorded administration date. The variable "GCSF before FN" indicates the administration of GCSF before chemotherapy and/or GCSF given during the chemotherapy period but before any occurrence of FN. Drug types used in chemotherapy were also included, and more information can be found in Supplementary Table S1.

Statistical Analyses
Descriptive analyses were conducted to characterize patient, tumor, and treatment characteristics. The associations between FN and the various characteristics were studied using the Chi-square test and Kruskal-Wallis test for categorical and continuous variables, respectively.
Univariable and multivariable Cox regression models were used to assess the association between the various risk factors and FN (survival package in R, where the Surv (time to event, FN event) command was used to estimate hazard ratios (HR) and corresponding 95% confidence intervals (CI)). For patients with FN, "time to event" was defined as the time from the start of chemotherapy to the first FN occurrence. For patients without FN, "time to event" was defined as the time from the start of chemotherapy to the end of chemotherapy. In view of the length of chemotherapy treatment courses, time to event was truncated at 210 days (6 months of chemotherapy + 30 days). Patients with missing values for continuous variables were not included in the univariate analysis. In the multivariable Cox regression model, the model was adjusted for all factors significantly associated with FN risk in the univariable model. Cox.zph was used to test for the proportionality of the Cox regression model fits. In addition, the Kaplan-Meier (KM) method was used to analyze FN risk by stage and GCSF status subsets (survival package in R); survival curves were compared using the log-rank test.
To assess patient, tumor, and treatment differences across ethnicities, multinomial logistic regression models were used, adjusting for age at diagnosis. Linear regression models for patients with complete records were also used to evaluate differences in ANCs across ethnicities, adjusted for age at diagnosis and BMI. A test of equal or given proportions for missingness in baseline BMI was conducted using prop.test. A Sensitivity analysis, including patients with missing baseline BMIs, was done.
Finally, to evaluate how different patient, tumor, and treatment characteristics might interact with ethnicity to modify the risk of FN, two-way interaction terms were created and assessed.  Table S2). Table 1 shows the descriptive statistics of patients' characteristics. Among the 7449 patients included in the JBCR, 6484 (87.0%) did not experience FN, and the remaining 965 (13.0%) were diagnosed with FN at least once during their chemotherapy treatment. The mean age at diagnosis was 54 years (IQR: 47.0 to 61.0), and the majority of the patients were of Chinese ethnicity (80.5%). The mean baseline BMI was 23.5 (IQR: 20.6 to 27.0) and the mean baseline ANC was 4.0 10 9 /L (IQR: 3.1 10 9 /L to 5.1 10 9 /L). The most frequently used chemotherapy regimen was taxanes in combination with anthracyclines and cyclophosphamides (54.2%). A total of 1545 (20.7%) patients were given GCSF before FN. Of the 965 first occurrences of FN, 497 (51.5%) occurred within the first month of chemotherapy ( Figure 2).

Ethnicity Is Associated with FN Occurrence after Adjusting for Patient, Tumor, and Treatment Characteristics
In univariate Cox models, those of non-Chinese ethnicity, diagnosed before 2008, older at diagnosis, greater maximum decrease in BMI, lower baseline ANC, negative ER status, triple-negative subtype, and anthracycline-containing regimens were significantly associated with FN occurrence ( Table 2, refer to Supplementary Table S3 for additional adjusted models). Additionally, patients who underwent additional treatments such as radiotherapy, herceptin, or tamoxifen treatments, had a significantly lower risk of developing FN. Further adjustments for patient, tumor and treatment variables that were significant in the univariate cox models did not result in any appreciable change in the hazard ratios and ethnicity remained as a significant factor contributing to FN risk (HRMalay:1.29 [1.07-1.54], p = 0.006; HRIndian:1.50 [1.19-1.88], p < 0.001). Table 2. Associations of variables with FN occurrence (n = 7449). Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using the Cox regression models. Values are considered statistically significant when p-value < 0.05.

Univariate
Adjusted

Ethnicity Is Associated with FN Occurrence after Adjusting for Patient, Tumor, and Treatment Characteristics
In univariate Cox models, those of non-Chinese ethnicity, diagnosed before 2008, older at diagnosis, greater maximum decrease in BMI, lower baseline ANC, negative ER status, triple-negative subtype, and anthracycline-containing regimens were significantly associated with FN occurrence ( Table 2, refer to Supplementary Table S3 for additional adjusted models). Additionally, patients who underwent additional treatments such as radiotherapy, herceptin, or tamoxifen treatments, had a significantly lower risk of developing FN. Further adjustments for patient, tumor and treatment variables that were significant in the univariate cox models did not result in any appreciable change in the hazard ratios and ethnicity remained as a significant factor contributing to FN risk (        Further subset analysis was done according to the presence or absence of GCSF before FN. Among those who received GCSF before FN, Malay patients displayed a higher risk for FN (HR Malay : 1.39 [1.02-1.88], p = 0.034; log-rank p: 0.053) (Figure 4a). On the other hand, among those who did not receive GCSF before FN, Indian patients showed a higher risk for FN (HR Indian : 1.56 [1.18-2.07], p = 0.002; log-rank p: 0.015) (Figure 4b).  Table 3 shows age-adjusted associations between ethnicity and various patient, tumor, and treatment characteristics in multinomial regression models. The result of the multinomial regression model can be interpreted as follows: the odds of a Malay patient developing stage III breast cancer is 2.3 times that of a Chinese patient. Apart from treatment institution and diagnosis year, patient characteristics were not significantly different between ethnic groups. Compared to the referent Chinese group, Malays, and Indians were more likely to be diagnosed with late-stage cancers and larger tumor sizes. When looking at proxy subtypes, Malays were at higher odds of developing luminal B (HER2neg) and HER2-overexpressed cancers, while Indians were at higher odds of developing triple-negative cancers. Additionally, among those not prescribed GCSF before chemotherapy, even though Indian patients have a higher baseline ANC than Chinese   Table S4).

Tumor and Treatment Characteristics Differ across Ethnicities
In terms of treatment, Malays and Indians were less likely to have undergone tamoxifen and radiotherapy. Additionally, Malays were more likely to have been administered anthracyclines but less likely to be prescribed GCSF before FN (Table 3).  Table 3 shows age-adjusted associations between ethnicity and various patient, tumor, and treatment characteristics in multinomial regression models. The result of the multinomial regression model can be interpreted as follows: the odds of a Malay patient developing stage III breast cancer is 2.3 times that of a Chinese patient. Apart from treatment institution and diagnosis year, patient characteristics were not significantly different between ethnic groups. Compared to the referent Chinese group, Malays, and Indians were more likely to be diagnosed with late-stage cancers and larger tumor sizes. When looking at proxy subtypes, Malays were at higher odds of developing luminal B (HER2-neg) and HER2-overexpressed cancers, while Indians were at higher odds of developing triple-negative cancers. Additionally, among those not prescribed GCSF before chemotherapy, even though Indian patients have a higher baseline ANC than Chinese (Beta Indian: 0.51 [0.29-0.73], p < 0.001), they also experienced a greater drop in ANC during chemotherapy (Beta Indian: 0.41 [0.16-0.67], p = 0.001) (    In terms of treatment, Malays and Indians were less likely to have undergone tamoxifen and radiotherapy. Additionally, Malays were more likely to have been administered anthracyclines but less likely to be prescribed GCSF before FN (Table 3). Table 4. Complete records analysis of associations between absolute neutrophil count (ANC) and ethnicity among patients without GCSF before the start of chemotherapy, adjusted for age at diagnosis, and baseline BMI. Beta values and 95% confidence intervals (CI) were estimated using the linear regression models. Values are considered statistically significant when the p-value < 0.05.

Various Tumor and Treatment Characteristics Did Not Differentially Affect FN Risk across Ethnicities
Additional models were constructed to further evaluate the interaction between ethnicity and other significantly associated variables in the univariable Cox models (Supplementary Table S5). No key interactions were observed.

Non-Chinese Breast Cancer Patients More Frequently Developed Multiple FN Episodes
We further investigated the total number of FN episodes that occurred during the chemotherapy treatment period. The age-adjusted multinomial model showed that both Malay (OR Malay [95%CI]: 3.65 [1.45-9.16], p = 0.006) and Indian patients (OR Indian : 4.43 [1.45-13.60], p = 0.009) were significantly more likely to develop more than 3 episodes of FN during treatment as compared to Chinese patients (Table 3).

Discussion
In the SGBCC population, comprising 1014 chemotherapy-treated breast cancer patients, the incidence rate of FN was 16.7%. Notably, patients of non-Chinese ethnicity had a higher FN incidence. The ethnic disparity was also seen in an independent JBCR hospital-register-based population of 7449 chemotherapy-treated breast cancer patients, where 13% of patients experienced FN episodes. In JBCR, non-Chinese were estimated to be at a 1.5 times elevated risk of developing chemotherapy-induced FN. Patients of non-Chinese ethnicity also more frequently experienced multiple FN episodes. Other significant factors associated with FN include age at diagnosis, anthracycline-containing regimens, and the absence of non-chemotherapy treatments, such as radiotherapy, herceptin, and tamoxifen treatment. While exploring differences across ethnicities, we found that Malays and Indians were more likely to be diagnosed at a later stage. These two groups were also less likely to undergo additional treatments, more likely to be prescribed with an anthracycline-containing regimen, and experienced greater dips in ANC during chemotherapy, which may contribute to the elevated FN risk. However, the ethnic disparity remained even after significantly associated patient, tumor, and treatment characteristics were adjusted for.
Previously reported patient-specific FN risk factors among Asian breast cancer patients include low BMI (BMI < 23 kg/m 2 ), low baseline white blood cell count, and low ANC [18][19][20]. In addition, age, poor performance status, and particular genetic variants have all been linked to elevated FN risk [20]. Similarly, our study showed that age and low baseline ANC were risk factors for FN.
Our study observed that anthracycline-containing regimens placed patients at higher risk for FN. However, this trend differs from what is commonly observed, where studies have shown that a taxane-based regimen contributes to higher FN incidence [21][22][23]. It should be noted that most of these studies focused on Caucasian populations, which may explain the differences in toxicities observed when compared to our Asian population. A Nigerian study comprising 113 breast cancer patients reported the lack of a significant relationship between taxane-based regimens and FN risk [24]. The authors attributed this finding to the fact that the majority of the patients were first started on anthracycline-based treatments, but 36% were later switched to a taxane-based regimen [24]. By chemotherapy regimen in Chen et al.'s study in Singapore, FN incidence observed in those treated with anthracycline-and taxane-based drugs was 15.0% and 11.7%, respectively [25], which concurs with our finding. Additionally, our study also found that compared to patients who received a combination of only anthracyclines and cyclophosphamides, those who received taxanes, anthracyclines, and cyclophosphamides were at lower risk for FN, even though this association was not significant. In agreement, a meta-analysis of 15 randomized trials by Zheng et al. reported that combined cytotoxic regimens led to significantly fewer FN events than single agents [26].
Ethnicity has been suggested to be associated with FN incidence previously [6][7][8][9]11,12]. Ethnic variance in treatment toxicity is widely acknowledged as a key element explaining inter-individual variation in cancer treatment responsiveness [13]. East Asians have been reported to be more susceptible to toxic side effects of chemotherapy than their Caucasian counterparts, requiring dose reduction in many studies [13]. In an earlier work by Chen et al. on 583 Asian cancer patients in Singapore, it was reported that FN incidence in Chinese, Malay, and Indian patients was 15.6%, 24.5%, and 25%, respectively [25]. In agreement, the results from our study showed that both Malay and Indian patients were at higher risk of FN, where FN incidence in Chinese, Malay, and Indian patients was 12.3%, 14.6%, and 18.1%, respectively.
In local clinical practice, patients diagnosed with locally advanced disease are prescribed anthracycline in combination with taxanes, while those with early-stage disease are given taxanes only. Similarly, the Malay patients in our group were observed to be diagnosed at a higher stage and were more likely to be given anthracycline in addition to taxanes, which can explain their greater FN risk. However, even after accounting for significantly associated patient, tumor, and treatment factors, the disparity between the effect of chemotherapy on FN risk persisted across ethnicities, indicating the possibility that other unidentified contributors exist. In our study, we found that Indian patients experienced larger dips in ANC during chemotherapy, which prompts further investigation in this area. Furthermore, inter-individual and ethnic differences in how docetaxel and doxorubicin are metabolized, as well as interethnic disparities in docetaxel-and doxorubicin-induced myelosuppression, are shown to impact the results of treatment [10,27,28]. Additionally, previous studies have shown that patients with comorbidities, such as renal and liver diseases, were at higher risk of systemic toxicities, such as FN [25,29]. This, combined with the widely observed trend that Malays and Indians are more likely to suffer from chronic conditions [30][31][32], can explain the higher FN incidence experienced. However, data on comorbidities were not properly collected for this study, which limited further investigation in this area.
This study found a consistent ethnic disparity in chemotherapy-induced FN in two large datasets comprising Chinese, Malay, and Indian breast cancer patients. However, several limitations should be noted. Due to the retrospective design of this study, selection bias cannot be eliminated. Another limitation is the incompleteness of clinical records, such as detailed information on comorbidities and chemotherapy regimens. The dataset has a sizable proportion of subjects (28%) with missing baseline BMI values. However, excluding patients with missing baseline BMI biased the results for absolute neutrophil count towards the null. Furthermore, commonly studied outcomes such as FN incidence during each cycle or FN risk due to differences in regimen dosage were unable to be studied due to the lack of information on the chemotherapy cycle or complete regimen details. Although efforts were made to adjust for several different treatment and demographic factors associated with FN, likely, there is still significant residual confounding in the results.

Conclusions
In summary, the results of this study suggest that there are ethnic differences in the risk of chemotherapy-induced FN. The findings present an opportunity to improve the precision of breast cancer treatment. Taking into consideration the susceptibility of certain ethnic groups to adverse effects, prophylactic or alternative treatments can be utilized to avoid these side effects. Moreover, close monitoring for early signs of infection may be emphasized in high-risk ethnic populations. Additional research where data for treatment regimens are properly collected is needed to investigate the possibility of pharmacogenetic differences across these ethnicities and formulate better-tailored treatment combinations. Further studies can also explore differences in the effects of FN across ethnicities, such as the need for reductions in relative dose intensity or delays in chemotherapy schedules, as well as survival outcomes.   Informed Consent Statement: Informed consent was obtained from all patients of SGBCC. They were informed that they can withdraw their consent and stop participation at any time without disclosing the reasons and without negative consequences for their future medical care. All patients have consented to publishing their data anonymously. As JBCR relies on deidentified registry-based data, the need to obtain informed consent was waived.
Data Availability Statement: SGBCC data is available to interested researchers upon request, with certain access limitations. All inquiries should be sent to the scientific steering committee of the Singapore Breast Cancer Cohort (SGBCC). For further information, please email the Principal Investigator, Mikael Hartman, at ephbamh@nus.edu.sg. A list of accessible data is available at https://blog.nus.edu.sg/sgbcc/for--researchers/ (accessed on 1 February 2023). The Joint Breast Cancer Registry dataset may be requested at: https://www.nccs.com.sg/research--innovation/data--on--request (accessed on 1 February 2023).