The immune system is critical for the development and management of malignant tumors. Lymphocyte is a major component of the immune system while lymphopenia was reported to be associated with poor prognosis in patients with breast cancer and other malignant cancers (1–4). However, lymphocytes are extremely radiosensitive (5, 6) and exposure to as low as 1 Gy of radiation can destroy mature circulating lymphocytes (7). Therefore, radiotherapy (RT) can damage the immune system and cause potential immunosuppression while it is applied for its role on killing malignant tumor cells.

Radiation induced lymphopenia (RIL) is common and known as a negative prognostic factor in many types of malignant solid tumors (8–15). In breast cancer, we have demonstrated that 60.5% patients had lymphopenia and 92.7% patients had some degree of reduction in peripheral lymphocyte count (PLC) after adjuvant RT (16). RIL or the ratio of nadir PLC and pretreatment PLC were also found to be a potential predictor for ipsilateral breast tumor recurrence or 5-year disease-free survival in breast cancer (17, 18). Meanwhile, several phase III clinical trials involving immunotherapy showed positive results which have significantly changed the treatment strategy for triple negative breast cancer (19, 20). The keynote 522 study (20) had enrolled previously untreated stage II or stage III triple-negative breast cancer and most of them would receive adjuvant RT concurrent with pembrolizumab after surgery. The effect of immunotherapy might be weakened by RIL since patients with low PLC had lower immune response rate while receiving checkpoint inhibitors (21). Therefore, it is important to establish a prediction model of RIL to improve RT plan of a better lymphocyte-sparing technique to reduce treatment-related lymphopenia.

Studies had revealed that radiation dose to the lymphatic system played an important role on RIL (6, 12). Our prior study also found that several radiation dosimetric factors (such as mean lung dose) were significant risk factors for lymphopenia after RT in patients with breast cancer (16). A simple RIL model based on radiation dose to the lymphatic system will be helpful clinically to improve RT plan and minimize lymphopenia. The lymphatic system is a complicated system composed of lymphoid organs such as lymph nodes/ducts, bone marrow and thymus, non-lymphoid organs such as lung and liver where lymphocytes reside, and circulation blood which transport lymphocytes to different parts of the body (22). Here we hypothesized that radiation dose to the lymphocytes in circulation blood to be critical for RIL because it was observed in breast and brain radiation patients while there is little lymphoid tissue in radiation of these regions (13, 23, 24). In addition, RIL is directly measured according to the number of lymphocytes in a unit volume of blood, and the active lymphocytes in circulation blood may be more radiosensitive than non-active lymphocytes in other parts of the lymphatic system (5, 6).

However, it is difficult to determine the radiation dose to the lymphocytes in circulation blood because they are moving targets. We have previously developed a model for the effective dose to the circulating immune cells (EDIC) for thoracic radiation based on planning dosimetric data, and demonstrated that EDIC was associated with both overall survival (OS) and local progression free survival (LPFS) after thoracic radiotherapy in non-small cell lung cancer (NSCLC) patients (25). This EDIC model has been validated externally in NSCLC (26) and esophageal cancer (15, 27), and further extended to abdominal radiation with consideration of other immune organs beyond circulating blood, including spleen, lymphatic ducts and bone marrows (22). Because the main irradiated blood-containing structures of breast cancer are similar to that of other thoracic malignancies, we hypothesized that this modeled EDIC may also be applicable to breast cancer. The purpose of this study is two folds: 1) to investigate whether EDIC is associated with RIL in patients with breast cancer; 2) to develop a normal tissue complication probability (NTCP) model for RIL by considering the EDIC as the dose to the circulating lymphocytes, similar as the conventional NTCP model for other normal structures (such as NTCP of radiation induced pneumonitis versus lung dose).

This study of 735 patients at the first time in breast cancer demonstrated a significant correlation between EDIC and RIL, and that both post/preRT PLC ratio and postRT PLC decreased linearly (R²<0.4) with increasing EDIC. The post/preRT PLC ratio had a better correlation with EDIC in patients with normal preRT PLC, and the postRT PLC appeared to have a better correlation with EDIC in patients with preRT lymphopenia. Comparing to the numerical correlation by the linear regression, EDIC fitted better into the sigmoid-shaped NTCP model for RIL (R²>0.96). The EDIC value of 50% incidence of grade-1 RIL was 1.2 Gy, the corresponding EDIC to reach 50% of grade-2 and grade-3 RIL was 2.1 and 3.7 Gy, respectively.

The significance of EDIC for RIL in breast radiation makes biological sense and is supported by radiation physics rationales. Lymphocytes are produced in bone marrow and thymus, and then travel through blood circulation into various functional sites. Radiation can directly kill circulating immune cells as they pass through the irradiated field. EDIC is the estimation of equivalent dose to the circulating immune cells in blood contributed by the irradiation of the blood-containing organs in thoracic radiation according to the mean dose to these organs (such as MLD, MHD and ITDV). In addition, the radiation dose fractionation effect is also considered in the EDIC model (25). EDIC has been reported to be associated with RIL in other thoracic malignancies such as lung and esophageal cancers (15, 25–27). As a type of thoracic cancers, breast cancer is similar to the NSCLC and esophageal cancer in terms of that its main irradiated blood-containing organs are lung and heart. Therefore, the EDIC model may be also applicable in breast cancer. However, the irradiation pattern or dose distribution in these organs may be quite different for breast cancer in comparison to that of the NSCLC and esophageal cancer. The irradiated lung and heart volumes in breast cancer are usually much smaller. According to the EDIC model derivation (25), the contribution of a blood-containing organ (such as lung) to EDIC is approximately expressed as B%*Mean Organ Dose (MOD) when A% is larger than 20% and fraction number is larger than 20, where MOD is the mean organ dose, A% is the percentage blood-current that flow into the organ, B% is the percentage blood-volume that present within the organ. This expression was derived based on the assumption that the blood flow through the organs in a serial pattern, and should be applicable for any blood-containing organs with any non-uniform dose distributions. The lung is composed of 5 lobes and each lobe may be approximately considered as having a serial blood-flow pattern with A%=20%. From the physics model perspective, the EDIC model can be applied in breast cancer radiation. Indeed, significance of EDIC on both postRT PLC and RIL of this study support the validity of this model in breast cancer.

It is interesting to note that when a patient had a normal preRT PLC, the post/preRT PLC ratio had better correlation with EDIC. This is consistent with the theory that when the preRT PLC is normal, the killing of circulation lymphocytes in blood by radiation plays the most important role in the reduction of PLC. However, 14.3% patients had preRT lymphopenia in this study, which was reported to be related with prior chemotherapy in our another study of 1012 patients with early or locally advanced breast cancer (28). Therefore, we also explored the relationship of EDIC and lymphopenia in this population. When a patient had preRT lymphopenia, the absolute postRT PLC was better associated with EDIC. This is consistent with previous report that when the preRT PLC is low, the regeneration of new lymphocytes from other parts of lymphatic system begins to play a role (12). This regeneration of new lymphocytes weakens the association between EDIC and post/preRT PLC ratio. The lower the preRT PLC, the larger the deviation from the correlation of EDIC and post/preRT PLC ratio. For 3 patients with grade-3 preRT lymphopenia, 2 of them were outliers, while 3 over 18 patients were outliers for grade-2 preRT lymphopenia, and 2 over 84 patients were outliers for grade-1 preRT lymphopenia in this study.

It is important to note that this study also provided a NTCP model of RIL based solely on EDIC. The NTCP model of each grade of lymphopenia fits well with the sigmoid-curved dose toxicity relationship, using EDIC for both binned data or binary data for each patient without binning. These two approaches achieved quite consistent results, and bootstrap validation also confirmed the similar results. These data suggest that EDIC is likely a true key dosimetric parameter, or at least a good surrogate of the true dosimetric parameter that directly impacts the RIL. EDIC might be used as a reference for clinician in dose prescription and treatment planning. With the guide of this EDIC model in lymphopenia, we may improve our treatment to limit RIL. For example, adjustment of beam energies, directions and number of beams may be able to reduce EDIC and hence limit RIL. In addition, according to the EDIC derivation (25), reducing the radiation fraction number n may also reduce EDIC through the term of (n/45)^1/2, which is contributed by blood-containing organs with a relatively small A% (such as A% <25%). Thus, hypofractionated treatments such as stereotactic body radiation therapy (SBRT) may reduce EDIC. Indeed, SBRT was reported to be associated with significantly less RIL than conventionally fractionated RT in locally advanced pancreatic cancer and breast cancer (18, 29, 30). Hypofractionated radiation in this study was also found to have less risk of RIL (Table 1).

This is the first study to validate the association of EDIC and RIL in patients with breast cancer. However, it should be noted that several other dosimetric factors were also reported to be significantly associated with RIL. These factors include lung V5 and heart V50 in lung cancer, brain V25 in glioma, and mean whole body dose in esophageal cancer (9, 12, 13, 31–33). In our prior study, lung V5 and volume modulated arc therapy rather than three-dimensional conformal technique were also risk factors for severe RIL (16). Although EDIC is a combination of several dosimetric parameters, it may not be the best model. A more comprehensive model with an optimal combination of other dosimetric parameters (such as lung V5) and likely including immune organs like bone marrow and thymus may turn out to be a better predictor than the EDIC. Adding other clinical factors may further improve the prediction of RIL in breast cancer. Further study is needed to determine these parameters and the best combination of them for better prediction. From a theoretic view point, EDIC is just an approximation of the dose to the immune cells in the circulating blood. The dose to the lymphatic stations, bone marrow and other lymphatic structures may also affect RIL. The regeneration of new lymphocytes may complicate the prediction. It should also be noted that the survival of this study is not mature yet. Future studies will be performed to explore the association of EDIC and other dosimetric parameters with the long-term treatment outcomes of patients.

This study demonstrated that EDIC was significantly associated with radiation induced lymphopenia (RIL) in breast cancer. Using EDIC as the dose variable, the risk of RIL can be predicted nicely by a conventional NTCP model. The corresponding EDIC to induce 50% of grade-1, grade-2 and grade-3 RIL was 1.2, 2.1 and 3.7 Gy, respectively. Should it be validated by external datasets, these number may be used as reference to guide radiation plan optimization and improve survivals for patients.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the University of Hong Kong-Shenzhen Hospital (# 2019 098). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

FC: primary investigator: overall study hypothesis, study design, data collection, data analysis, result interpretation, manuscript writing and final manuscript approval; J-YJ and TH: study design, physics modeling, data analysis, manuscript writing and final manuscript approval; HZ, LM, and PF: various parts of statistical analysis, results interpretation and final manuscript approval; YH, FY, QC, YZ: clinic data collection and final manuscript approval; YN: radiation dosimetric factors collection and final manuscript approval; HJ, WW, LY, and ZX: results interpretation and final manuscript approval; F-MK: study idea, study hypothesis/design, data quality control, data analysis, result interpretation, detailed manuscript preparation and final manuscript approval. All authors contributed to the article and approved the submitted version.

This project was partially supported by Shenzhen Key Medical Discipline Construction Fund (No. SZXK014), Shenzhen Science and Technology program (Grant No: KQTD20180411185028798), Health Commission of Guangdong Province, China (A2021114) and a Varian Medical System research grant.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.