Immunologic Consequences of Sequencing Cancer Radiotherapy and Surgery.

SPECIAL SERIES: MATHEMATICAL ONCOLOGY original report Immunologic Consequences of Sequencing Cancer Radiotherapy and Surgery Juan Carlos L´ opez Alfonso, PhD 1 ; Jan Poleszczuk, PhD 2 ; Rachel Walker, PhD 3 ; Sungjune Kim, MD, PhD 3 ; Shari Pilon-Thomas, PhD 3 ; Jose J. Conejo-Garcia, MD, PhD 3 ; Hatem Soliman, MD 3 ; Brian Czerniecki, MD 3 ; Louis B. Harrison, MD 3 ; and Heiko Enderling, PhD 3 abstract PURPOSE Early-stage cancers are routinely treated with surgery followed by radiotherapy (SR). Radiotherapy before surgery (RS) has been widely ignored for some cancers. We evaluate overall survival (OS) and diseasefree survival (DFS) with SR and RS for different cancer types and simulate the plausibility of RSand SR-induced antitumor immunity contributing to outcomes. MATERIALS AND METHODS We analyzed a SEER data set of early-stage cancers treated with SR or RS. OS and DFS were calculated for cancers with sufficient numbers for statistical power (cancers of lung and bronchus, esophagus, rectum, cervix uteri, corpus uteri, and breast). We simulated the immunologic consequences of SR, RS, and radiotherapy alone in a mathematical model of tumor-immune interactions. RESULTS RS improved OS for cancers with low 20-year survival rates (lung: hazard ratio [HR], 0.88; P=.046) and improved DFS for cancers with higher survival (breast: HR = 0.64; P,.001). For rectal cancer, with intermediate 20-year survival, RS improved both OS (HR = 0.89; P= .006) and DFS (HR = 0.86; P=.04). Model simulations suggested that RS could increase OS by eliminating cancer for a broader range of model parameters and radiotherapy-induced antitumor immunity compared with SR for selected parameter combinations. This could create an immune memory that may explain increased DFS after RS for certain cancers. CONCLUSION Study results suggest plausibility that radiation to the bulk of the tumor could induce a more robust immune response and better harness the synergy of radiotherapy and antitumor immunity than postsurgical radiation to the tumor bed. This exploratory study provides motivation for prospective evaluation of immune activation of RS versus SR in controlled clinical studies. Clin Cancer Inform. © 2019 by American Society of Clinical Oncology INTRODUCTION Surgery followed by radiotherapy (SR) improves locoregional control in numerous locally advanced cancers, 1-3 making radiotherapy a vital component of intent-to-cure cancer treatment. 4 Radiotherapy followed by surgery (RS) versus surgery alone has demonstrated improved outcomes in bladder, cervical, esophageal, rectal, and resectable pancreatic cancers. Clinical trials comparing SR and RS outcomes are scarce and have been limited to a few cancer types. One successful example is rectal carcinoma, 5,6 where improved local control after RS versus SR (with or without concurrent chemotherapy) was demonstrated and resulted in a paradigm shift from SR to RS. However, these trials have failed to demonstrate superior overall survival (OS) after RS. Similarly, in a prospective trial for muscle-invasive bladder cancer, 7 no significant difference in 3-year statistics between RS and SR was found. A slight increase in OS for patients undergoing RS compared with SR was observed in soft tissue sarcoma of the limbs. 8 To date, the few published studies comparing radiotherapy and surgery sequencing seem inconclusive. A SEER analysis showed that RS can be associated with a significantly higher risk of death in Siewert type II gastroesophageal junction cancer when compared with SR. 9 However, for locally advanced esophageal cancer, and squamous cell carcinoma in particular, superiority of RS over SR with concurrent chemotherapy was demonstrated. 10 Retrospective analyses have also shown superior OS after RS in stage III non–small-lung cancer 11 and in cT2 noninflammatory breast cancer with concurrent chemotherapy. 12 There are also some indications that RS can reduce the risk of local recurrence in localized soft tissue sarcoma. 13 It is increasingly appreciated that radiotherapy increases the mutational burden and induces cell stress ASSOCIATED CONTENT Appendix Data Supplement Author affiliations and support information (if applicable) appear at the end of this article. Accepted on November 27, 2018 and published at ascopubs.org/journal/ cci on April 9, 2019: DOI https://doi.org/10. 1200/CCI.18.00075 1 Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. as well as immunogenic cell death, thereby exposing a wealth of previously hidden and de novo tumor-associated antigens, stress proteins, and danger-associated molecular patterns to the immune system. 14-18 Tumor-infiltrating lymphocyte enrichment after radiotherapy was previously assessed in 40 patients with rectal cancer. The densities of CD3 + and CD8 + T-lymphocytes significantly increased from preradiotherapy biopsy specimens to postradiotherapy surgically resected specimens. 19 It is conceivable that immune-related benefits of radiation to the bulk tumor compared with radiation to the postsurgical cavity could be general phenomena even in early-stage disease. Herein, we evaluate OS and disease-free survival (DFS) outcome data for various cancers and present a quantitative framework to simulate tumor–immune system dynamics during the different treatment sequences, which further supports the plausibility of increased antitumor immunity with RS compared with SR. MATERIALS AND METHODS SEER Analysis We queried the SEER database (November 2016 submission with additional treatment information) for localized cancers (no lymph node involvement and no metastasis; N0M0) with at least 100 cases of each of RS and SR and available information about potentially confounding covariates (Fig 1A). These included cancers of the lung and bronchus, esophagus, rectum, cervix uteri, corpus uteri, and breast. We selected patients who received either RS or SR and censored the follow-up period to 20 years. DFS was calculated as previously discussed. 19a Hazard ratios (HRs) of DFS and OS adjusted for age, sex, year of diagnosis, histology, type of surgery, tumor size, and treatment sequencing (RS vSR) were calculated using multivariable Cox proportional hazard models. 20 Modeling of Tumor–Immune System Dynamics During SR and RS Mathematical modeling of tumor–immune system interactions has a long history. 21,22 The established Kuznetsov model captures first-order principles underlying the complex disease dynamics and has been calibrated against experimental data. 21 This model relies on the following main assumptions: tumors follow logistic growth dynamics, where initially exponential growth decelerates as the tumor approaches the tissue-carrying capacity; tumor growth is modulated by the cytotoxic action of immunocompetent effector T cells as part of specific adaptive immune responses; cytotoxic effector T cells are recruited in response to tumor burden following Michaelis-Menten dynamics; tumor-infiltrating effector T cells may get exhausted by their antitumor activity and undergo spontaneous death; and innate immunity or base immune surveillance is represented as a baseline presence of effector T cells at any time, even in the absence of tumor cells. We extended the Kuznetsov model to explore the potential benefits and immunologic consequences of radiotherapy and surgery sequencing, on the basis of the following additional assumptions: radiotherapy kills both cancer and effector T cells 23,24 at different rates determined by the radiation dose and cell type–specific radiosensitivity by some forms of cell death, including autophagy, apoptosis, necrosis, mitotic catastrophe, and senescence; irradiated tumor cells may undergo immunogenic cell death that emits immunostimulating signals and chemokines, resulting in the recruitment of inflammatory immune cells to the tumor microenvironment, including antigen-presenting cells such as dendritic cells and macrophages, which in turn recruit and activate cytotoxic effector T cells 15,25-28 ; the strength of radiotherapy-induced antitumor response, characterized by subsequent immune cell infiltration of the tumor, is assumed to depend on the number of tumor cells killed by radiotherapy; and surgical resection is simulated by instantaneously decreasing both cancer and effector cell populations. We denoted the number of viable cancer cells (or clonogens) in an arbitrary tumor volume as C(t), tumorinfiltrating activated immune effector T cells as E(t), and irradiated cancer cells undergoing immunogenic cell death as D(t). The model was formulated as a system of ordinary differential equations given by dC/dt = (tumor growth) − (death of tumor cells by effector T cells) −(radiotherapy) − (surgical resection), or dC dt rC1−C+D k−apEC −δτiTRt,d,C −δτjTSt,C(1) by dE/dt = (tumor burden–stimulated effector T-cell recruitment) −(exhaustion of effector T cells by their antitumor action) + (physiologic level and decay of effector T cells) + (radiotherapy-induced immunostimulation) − (radiotherapy) −(surgical resection), or dE dt fC g+CE−a1−pEC +hEp−E+qD −δτiTRt,d,E−δτjTSt,E(2) and by dD/dt = (tumor cell sterilized by radiotherapy) − (lysis of tumor cells sterilized by radiotherapy), or dD dt δτiTRt,d,C−nD (3) where the time coordinate on the system variables was omitted for notational simplicity. δ(τ)istheDiracdelta function. For demonstration purposes, we chose previously reported specific model parameter values that simulate a growing tumor and associated effector cell dynamic in the L´ opez Alfonso et al 2© 2019 by American Society of Clinical Oncology Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. absence of therapy (Table 1). For analysis, we changed model parameters that may represent the patient-specific biology of individual patients, including different human leukocyte antigen haplotypes or increased mutational burden for different cancers that may translate into higher immunogenicity. The cytotoxic effect of radiotherapy on cancer cells was simulated using the linear quadratic (LQ) model that is A Records of patients with cancer of breast, cervix uteri, corpus uteri, rectum, esophagus, or lung and bronchus (N = 3,483,681) Excluded in total (n = 3,254,384) (n = 583,418) (n = 1,851,801) (n = 708,477) (n = 7) (n = 1,028) (n = 28,403) (n = 5,273) (n = 3,706) (n = 72,271) Prior cancer Regional or distant disease Did not receive RT in adjuvant or neoadjuvant setting Unknown age at diagnosis Unknown race Incomplete dates Missing data on second primary Unable to calculate cancer-free survival Unknown/undefined value of ≥ 1 covariate in multivariable analysis B 5101520 Follow-Up (years) Survival (probability) 0 0.25 0.50 0.75 1.00 Breast Cervix uteri Corpus uteri Rectum Esophagus 203,151 124,384 59,965 19,284 5,380 2,769 1,461 738 252 81 11,940 5,709 2,652 1,068 355 8,136 4,316 1,767 480 129 795 249 107 31 3 2,506 752 251 61 15 No. at risk: Lung and bronchus Surgery followed by radiation (SR) Breast Cervix uteri Corpus uteri Rectum Esophagus Lung and bronchus (n = 221,231) (n = 201,631) (n = 2,444) (n = 11,745) (n = 3,142) (n = 189) (n = 2,080) Breast Corpus uteri Rectum Lung and bronchus Cervix uteri Esophagus Radiation followed Breast Cervix uteri Corpus uteri Rectum Esophagus Lung and bronchus (n = 8,066) (n = 1,520) (n = 325) (n = 195) (n = 4,994) (n = 606) (n = 426) by surgery (RS) 0.56 0.54 0.37 0.30 0.13 0.04 FIG 1. (A) SEER inclusion/exclusion. (B) Kaplan-Meier 20-year survival. Sequencing Cancer Radiotherapy and Surgery JCO Clinical Cancer Informatics 3 Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. widely used in clinical applications 32-34 to approximate the dose-dependent surviving fraction SF(d) of irradiated cancer cells by SFde−ξ(αd+βd2)(4) where d(Gy) is the radiation dose and α(Gy −1 ) and β(Gy −2 ) are cell type–specific radiosensitivity parameters. Compelling evidence demonstrates that hypoxic (poorly oxygenated) cancer cells are growth arrested and estimated to be approximately three times more resistant to radiation than normoxic cycling cells. 30,35 We set ξ=1orξ= 1/3 to respectively scale the radiosensitivity of proliferative and quiescent cancer cells as previously demonstrated. 30,36 We set radiosensitivity parameters α=0.3Gy −1 and β=0.03Gy −2 as conventionally assumed and estimated for a variety of tumor types. 29 The probabilities of surviving 2 Gy radiation were SF (2 Gy) = 0.49 for proliferating cancer cells (ξ= 1) and SF (2 Gy) = 0.79 for quiescent cancer cells (ξ=1/3). Effector T-cell radiosensitivity was estimated on the basis of experimental data of radiotherapy-induced apoptosis in lymphocytes obtained from blood samples. 37 Apoptosis is considered one of the dominant cell death processes in response to radiotherapy, 38 and correlation between the intensity of apoptosis in lymphocytes and radiation dose has been reported. 37,39 From doseresponse curves of effector CD8 + T cells in vitro after exposure to acute doses of 0 to 8 Gy, 37 we derived SF E (2 Gy) = 0.61. Radiotherapy was simulated with a total dose of 50 Gy delivered in 25 weekday fractions of d= 2 Gy per day. The total dose of 50 Gy was chosen for demonstration purposes in line with the standard dose for breast cancer, the largest patient cohort in the SEER data. Extension of the analysis to other total doses was straightforward and did not alter the results of this study (Appendix Fig A1). After logistic tumor growth and the LQ model in Equation 4, 32,40 the fraction of cancer cells (C) sterilized by radiotherapy with dose dat time τ i was determined by the loss term T R (τ i ,d,C)in Equation 1 given by TRτi,d,C1−SFpdCτi1−Cτi+Dτi k +1−SFqdCτiCτi+Dτi k (5) where the terms C(τi)1−C(τi)+D(τi) kand C(τi)C(τi)+D(τi) kare the number of proliferating and quiescent cancer cells at time τ i such that their sum is equal to C(τ i ). The radiobiologic terms 1 −SF p (d) and 1 −SF q (d) represent respectively the fraction of proliferating and quiescent cancer cell sterilized by radiotherapy with dose dand determined by the LQ model in Equation 4. TABLE 1. Overview of Model Parameters and Values Used in Simulations Parameter Description Value Unit First Author rTumor growth rate (1.0 - 4.0) ×10 −1 Day −1 Poleszczuk 22 kTumor tissue carrying capacity 5.32 ×10 8 Cells Poleszczuk 22 aT cell–cancer cell interactions constant 0.14 ×10 −6 Day −1 cells −1 Poleszczuk 22 pProbability that during T cell–cancer cell interactions the latter is killed 9.98 ×10 −1 Dimensionless Poleszczuk 22 fMagnitude of immune system stimulation by the presence of cancer cells (1.0 - 3.0) ×10 −1 Day −1 Poleszczuk 22 gImmune stimulation damping coefficient 0.16 ×10 6 Cells Poleszczuk 22 hEffector cells decay rate 5.9 ×10 −1 Day −1 Poleszczuk 22 E* Physiologic level of effector cells 0.3 ×10 6 Cells Poleszczuk 22 αRadiation sensitivity parameter 0.3 Gy −1 Guerrero 29 βRadiation sensitivity parameter 0.03 Gy −2 Guerrero 29 ξCancer cell radiation resistance 1 (proliferating cells) Dimensionless Enderling 30 1/3 (quiescent cells) qRT-induced immunostimulation (0.0 −6.5) ×10 −1 Day −1 Assumed for demonstration purpose nLysis rate of tumor cells killed by RT 6.9 ×10 −1 Day −1 Eftimie 31 Abbreviation: RT, radiotherapy. L´ opez Alfonso et al 4© 2019 by American Society of Clinical Oncology Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. The loss term T R (τ i ,d,E) in Equation 2 simulates the killing effects of radiotherapy on effector T cells (E) at time τ i and is given by TRτi,d,E1−SFEdEτi(6) Surgical resection was simulated by instantaneously decreasing both cancer and effector cell populations by 99%. At the time of surgery τ j , removal of cell populations was incorporated in the model by the following loss terms TSτj,CλCCτj(7) TSτj,EλEEτj(8) where surgical resection was simulated with λ C =λ E = 0.99. After surgical resection, the dynamics of remaining cell populations, if any, continued to be governed by Equations 1 to 3. Model simulations were initialized with 10 6 viable cancer cells (or clonogens) and 10 5 effector T cells in an arbitrary tumor volume. Tumor–immune system interaction dynamics were simulated for different combinations of tumor growth rate (r) and recruitment of effector T cells in response to tumor burden (f). Combinations of parameters r and fthat yielded an increase of cancer cells .10 8 were considered for treatment simulations. Note that this number represents viable cancer cells in a tumor that may have considerably larger radiographic volumes. Tumor control (TC) was assumed if the number of viable cancer cells dropped below C(t) ,1. The mathematical model was simulated using Matlab (www.mathworks.com). RESULTS OS and DFS of RS and SR Cancer sites were evaluated for 20-year OS (Fig 1B)and DFS (Appendix Fig A2). A trend of RS benefit versus SR became apparent when cancers were sorted by their 20-year OS (DFS regression coefficient, −0.69; P,.05; OS regression coefficient, +0.35; P,.05; Fig 2). For cancers of the lung and bronchus (n = 2,506), RS yielded significantlyimprovedOS(HR=0.88;P= .046; 20-year survival, 6.2% v2.5%; Fig 2B). Although an OS benefit was not observed in breast cancer (n = 203,151) with high baseline 20-year OS rates, RS yielded a significant improvement in DFS (HR = 0.64; P,.001; Fig 2A). Interestingly, patients with rectal cancer (n = 8,136) with an intermediate 20-year survival rate of 30% may benefit from RS, with both increased OS (HR = 0.89; P= .006) and DFS (HR = 0.86; P= .04). Crucially, RS was not found to be significantly inferior to SR in any of the multivariable Cox proportional model analyses (Data Supplement). Analysis of propensity score–matched cohorts is shown in Appendix Figure A3. Effects of Treatment Order Negligible Without Simulated Radiotherapy-Induced Immunity The effectiveness of radiotherapy in eliminating cancer may result from a combination of the lethal effect of radiotherapy and stimulation of antitumor immunity. We first simulated treatments without additional radiotherapy-induced immunity (q= 0 day −1 in Equation 2). Simulations of SR and RS yielded similar outcomes, with tumor control (TC) being dependent on tumor growth rate and pretreatment tumor size. Interestingly, radiotherapy alone could not eliminate Lung and bronchus Rectum HR = 0.86 P = .04 Breast HR = 0.64 P < .001 Corpus uteri Esophagus Cervix uteri 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0 102030405060 20-Year OS (%) DFS HR (neoadjuvant v adjuvant RT) A HRDFS = –0.69* DFS20y + 1.08† 20-Year OS (%) Lung and bronchus HR = 0.88 P = .046 Rectum HR = 0.89 P = .006 Breast Corpus uteri Esophagus Cervix uteri 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0 102030405060 OS HR (neoadjuvant v adjuvant RT) B HROS = 0.35* OS20y + 0.84† FIG 2. Hazard ratios (HRs; with 95% CIs, Pvalues, and linear least squares regression lines weighted by the inverses of site-specific CIs) for (A) disease-free survival (DFS) and (B) overall survival (OS) after preoperative (neoadjuvant) RT (RS) compared with postoperative (adjuvant) RT (SR). HRs were adjusted for age, sex, year of diagnosis, histology, type of surgery, type of applied radiation, and tumor size. Equation describes the trend line, and coefficients were tested for significant difference from 0 using ttest. (*) P,.05. (†)P,.001. RT, radiotherapy Sequencing Cancer Radiotherapy and Surgery JCO Clinical Cancer Informatics 5 Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. tumors with as few as 10 7 pretreatment viable cancer cells (Fig 3). RS Improves Outcomes With Radiotherapy-Induced Immunity Because radiotherapy-induced cytotoxicity alone was unlikely to eradicate all cancer cells with the chosen radiosensitivity parameters, but radiotherapy often sterilized large tumors with presumably .10 7 viable cancer cells, clinically observed complete tumor eradication may depend on the strength of radiotherapy-induced antitumor immunity activation (q, day −1 in Equation 2). The efficacy of radiotherapy alone and RS but not SR increased with increased radiotherapy-induced immunity (q=0.25to 0.65 day −1 ). In addition, for strong radiotherapy-induced antitumor immunity, radiotherapy alone resulted in better outcomes than SR. RS yielded TC for more model parameter combinations than radiotherapy alone and SR (Fig 4). For SR, increased radiotherapy-induced immunostimulation had no notable impact on outcomes, because a majority of tumor cells were removed preirradiation, and radiotherapy-induced antitumor immunity was not sufficiently achieved. The benefitofradiotherapyalone and RS over SR was independent of the total number of radiation fractions (Appendix Fig A1) and insensitive to changes in other model parameters (Appendix Figs A4 to A6). RS Induces Stronger Antitumor Immunity Than SR After surgery, radiation is delivered with the intent to eradicate residual cancer cells or microscopic tumors in the tumor bed beyond the surgical margins. Compared with SR RSRT alone TC by treatment PD 1.0 × 107 1.0 × 108 Pretreatment Viable Cancer Cells C (t) A B 108 106 104 102 100 No. of Cells –17 0 34 68 102 Time (days) (RT) 108 106 104 102 100 No. of Cells –20 –10 0 10 20 30 Time (days) (SR) 108 106 104 102 100 No. of Cells –20 –10 0 10 20 30 Time (days) (RS) r (day–1) f (day–1)f (day–1) f (day–1)f (day–1) f (day–1)f (day–1) 0.3 0.2 0.1 0.3 0.2 0.1 0.1 0.2 0.3 0.4 0.3 0.2 0.1 0.3 0.2 0.1 TC PD r (day–1)r (day–1) r (day–1)r (day–1)r (day–1) 0.1 0.2 0.3 0.4 0.3 0.2 0.1 0.3 0.2 0.1 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4 0.1 0.2 0.3 0.4 RTViable cancer cells C(t) Effector cells E(t) FIG 3. (A) Model-predicted treatment outcomes of radiotherapy (RT) alone, RT after surgical resection (SR), and RT before surgical resection (RS) without RT-induced immunity (ie, q= 0.0 day −1 in Equation 2) in a cohort of virtual patients with different combinations of pretreatment tumor size, tumor growth rate (r), and immune recruitment rate (f) in response to tumor burden. Radiation is delivered to a total dose of 50 Gy in 25 daily fractions at 2 Gy per day, 5 days per week. Tumor control (TC) by treatment (blue) and progressive disease (PD; red) refer to tumor eradication and escape after treatment, respectively. (B) Time evolution of tumor and effector T cells corresponding to the location marked by stars in panel A. L´ opez Alfonso et al 6© 2019 by American Society of Clinical Oncology Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. SR RSRT alone TC by treatment PD 2.5 × 10–1 4.5 × 10–1 RT-Induced Immunostimulation q (day –1) A B 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells 108 106 104 102 100 No. of Cells –27 0 27 54 81 108 Time (days) (RT) –27 0 27 54 81 108 Time (days) (SR) –27 –14 0 14 27 Time (days) (RS) –27 –14 0 14 27 Time (days) (RT) –27 0 27 54 81 108 Time (days) (SR) –27 –14 0 14 27 Time (days) (RS) –17 0 17 34 6851 Time (days) (RT) –170 17345168 Time (days) (SR) –17 0 17 34 Time (days) (RS) r (day–1) f (day–1)f (day–1)f (day–1) f (day–1)f (day–1)f (day–1) f (day–1)f (day–1)f (day–1) 0.3 0.2 0.1 0.3 0.2 0.1 0.1 0.2 0.3 0.4 6.5 × 10–1 0.3 0.2 0.1 0.3 0.2 0.1 TC PD r (day–1) 0.1 0.2 0.3 0.4 0.3 0.2 0.1 0.3 0.2 0.1 r (day–1) 0.1 0.2 0.3 0.4 r (day–1) 0.1 0.2 0.3 0.4 r (day–1) 0.1 0.2 0.3 0.4 r (day–1) 0.1 0.2 0.3 0.4 RT Effector cells E(t) Viable cancer cells C(t) r (day–1) 0.3 0.2 0.1 0.1 0.2 0.3 0.4 0.3 0.2 0.1 r (day–1) 0.1 0.2 0.3 0.4 0.3 0.2 0.1 r (day–1) 0.1 0.2 0.3 0.4 Sequencing Cancer Radiotherapy and Surgery JCO Clinical Cancer Informatics 7 Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. RS, SR implied a significantly smaller number of cancer cells for radiotherapy-induced immune activation. The integral of the term qD in Equation 2 simulated the number of effector cells recruited to the tumor. Simulation analysis confirmed that the strength of tumor-specific immunity induced by RS was significantly higher than that induced by SR (Figs 5A and 5B). This may contribute to both increased OS rates and increased antitumor immunity, which may surveil metastatic deposits outside the irradiation field (abscopal response) to increase DFS (compare with Fig 2). Radiotherapy-Induced Immunity Suggests Plausibility of RS Dose De-escalation Normal tissue toxicity remained a limiting factor in radiotherapy, with acute and late toxicities largely being attributable to total radiation dose and dose per fraction. Model simulations suggested that for high rates of radiotherapyinduced immunity, RS with 2 Gy per fraction may be reduced to as low as half the number of RS fractions for outcomes comparable to the full course of SR (12-fraction RS v25-fraction SR; Fig 5C). Lower immunogenicity may require .13 fractions but potentially provides TC probability with fewer fractions than SR. For high rates of radiotherapy-induced immunity, radiotherapy alone may be a potential alternative to SR, with comparable TC rates (Fig 5D). DISCUSSION Accumulating evidence supports the notion that radiotherapy-induced antitumor immunity, especially when promoted with concurrent immunotherapy, could result in long-term memory 41,42 and may yield robust systemic immune responses against metastatic reservoirs outside the irradiation field, a phenomenon termed the abscopal effect. 43-45 Radiotherapy may no longer be considered a purely cytotoxic treatment; rather, it has biologic properties to transform the tumor into an in situ vaccine. 46 The clinical promise of the complex but promising immuneradiation synergy is reflected in .100 active clinical trials testing the combination of radiotherapy with immunotherapy, mostly for metastatic cancer. 47 A robust immune response after focal irradiation may require a sufficiently large population of cancer cells to activate immune cells in numbers large enough to disseminate systematically and establish an immunologic memory. The herein presented model simulations suggest a positive correlation between radiotherapy-induced antitumor immunity and tumor size at time of irradiation and may explain the poorer local outcomes achieved with SR compared with RS. After surgical resection with clean margins, radiotherapy will, at best, act upon microscopic tumor clusters in the tumor bed, which may be insufficient to elicit a robust immune response. For cancers with poor OS rates even when detected early (cancers of lung and bronchus, esophagus, rectum), RS may increase OS compared with SR because of additional cytotoxicity resulting from the increased immune activation. For tumors with higher OS rates (breast, rectum), benefits of RS can be seen in DFS, arguably because of higher antitumor immunity that may also act systemically against tumors outside the irradiation field and future disease. 22 For some tumors, no significant outcome benefits were observable for either sequencing of radiotherapy and surgery. However, trends toward increased OS for cancers with low survival rates (regression coefficient, −0.65; P,.05) as well as increased DFS for cancers with high survival rates (regression coefficient, 0.35; P,.05) were noted. One inherent limitation of SEER data analysis is composed of the potentially confounding factors that are not recorded, including patient selection criteria. Patients eligible for RS may be disproportionally treated in clinical trials and thus could represent a substantially different population than patients undergoing SR. Moreover, RS may be preferentially performed by academic, high-volume centers that often provide better outcomes in general. To address the issue of unbalanced arms, we performed multivariate Cox proportional hazards model analysis after preprocessing the data with propensity score matching on the basis of all considered covariates. 48 These data are comparable to the results obtained for the whole cohorts (Appendix Fig A3). Different mathematical models have been proposed to simulate tumor–immune system interactions 21,22,49-59 and the synergistic effects of radiotherapy with the immune system and immunotherapy. 22,60-63 To demonstrate immunologic consequences of sequencing radiotherapy and surgery, we extended the established Kuznetsov model and its parameterization as previously discussed. 21,22 Model simulations revealed that observable clinical responses may be parameter dependent, and as such, the model needs to be rigorously calibrated for individual cancers before definite conclusions can be drawn. Taken together, the statistical analysis of historical outcome data and the mathematical model combine to further add to the growing interest in understanding the biologic and immunologic consequences of radiotherapy. This study FIG 4. (A) Model-predicted treatment outcomes of radiotherapy (RT) alone, RT after surgical resection (SR), and RT before surgical resection (RS) for tumors of 10 8 viable cancer cells pretreatment and increasing strength of RT-induced immunostimulation (q, day −1 in Equation 2) in a cohort of virtual patients with different combinations of tumor growth rate (r) and immune recruitment rate (f) in response to tumor burden. Radiation is delivered to a total dose of 50 Gy in 25 daily fractions at 2 Gy per day, 5 days per week. Tumor control (TC) by treatment (blue) and progressive disease (PD; red) refer to tumor eradication and escape after treatment, respectively. (B) Time evolution of tumor and effector T cells corresponding to the locations marked by diamonds, triangles, and stars in panel A. L´ opez Alfonso et al 8© 2019 by American Society of Clinical Oncology Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. provides rationale and motivation for additional investigating and understanding the effects of radiotherapy on the immune system of patients with cancer to guide appropriate prospective validation of radiotherapy and surgery sequencing. This should include immune panels on longitudinal blood draws during radiotherapy as well as analysis of immune infiltration in preirradiation biopsies and surgically resected tissues. AFFILIATIONS 1 Braunschweig Integrated Centre of Systems Biology, Helmholtz Centre for Infection Research, Braunschweig, Germany 2 Nalecz Institute of Biocybernetics and Biomedical Engineering, Polish Academy of Sciences, Warsaw, Poland 3 Integrated Mathematical Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL CORRESPONDING AUTHOR Heiko Enderling, PhD, Department of Integrated Mathematical Oncology, H. Lee Moffitt Cancer Center and Research Institute, 12902 Magnolia Drive, SRB 4, Tampa, FL 33612; e-mail: heiko.enderling@moffitt.org. EQUAL CONTRIBUTION J.C.L.A. and J.P. contributed equally to this work. SUPPORT Supported in part by Institutional Research Grant No. 14-189-19 from the American Cancer Society (H.E.), by the German Federal Ministry of Education and Research funding of the eMED project SYSIMIT (Grant No. 01ZX1308D; J.C.L.A.), and by the Helmholtz Association of German Research Centers Initiative and Networking Fund for the project on reduced complexity models (Grant No. ZT-I-0010; J.C.L.A.). AUTHOR CONTRIBUTIONS Conception and design: Juan Carlos L´ opez Alfonso, Jan Poleszczuk, Shari Pilon-Thomas, Brian Czerniecki, Louis B. Harrison, Heiko Enderling Financial support: Juan Carlos L´ opez Alfonso, Heiko Enderling Collection and assembly of data: Jan Poleszczuk, Brian Czerniecki 1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 Parameter Space Controlled (%) Parameter Space Controlled (%) 1 5 10 15 20 2515 10 No. of Treatment Fractions No. of Treatment Fractions 15 20 25 CD SR 2 Gy × 25 fractions q(day–1) 4.5 × 10–1 2.5 × 10–1 6.5 × 10–1 SR 2 Gy × 25 fractions TC by treatment (SR) 0.1 0.2 0.3 0.4 0.3 0.2 0.1 0.3 0.2 0.1 ATC by treatment (RS) r (day–1) f (day–1) f (day–1) r (day–1) 0.1 0.2 0.3 0.4 3 × 107 1 × 107 3 × 106 1 × 106 No. of Effector Cells B FIG 5. Comparison of immune recruitment and treatment outcomes for tumors of 10 8 viable cancer cells pretreatment with tumor growth rate (r) between 0.1 and 0.4 day −1 and immune recruitment rate (f) in response to tumor burden between 0.1 and 0.3 day −1 . (A) Recruitment of effector T cells to the tumor bed as a result of radiotherapy (RT) –induced immune responses by surgery followed by RT (SR). (B) Recruitment of effector T cells to the tumor bed as a result of RT-induced immune responses by RT followed by surgery (RS). Recruitment of effector T cells after both SR and RS is estimated by the integral of the term qD in Equation 2, with q= 4.5 ×10 −1 day −1 . Arrows point to the parameter combinations that yield tumor control (TC) by SR and RS (c.f. Fig 4A). (C) Proportion of parameter space controlled for SR (25 daily fractions at 2 Gy per day, 5 days per week; blue line) compared with RS with gradually increasing number of fractionations. (D) Proportion of parameter space controlled for SR (25 daily fractions at 2 Gy per day, 5 days per week; blue line) compared with RT alone with increasing number of fractionations. Parameters are 0.1 ≤r≤0.4 day −1 and 0.1 ≤f≤0.3 day −1 . Different strengths of RT-induced immunostimulation are color coded (q, day −1 in Equation 2). Sequencing Cancer Radiotherapy and Surgery JCO Clinical Cancer Informatics 9 Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved. SR RSRT alone Surviving Fraction of Effector Cells at 2 Gy (SF 2 ) 0.410.610.81 TC by treatment PD r (day –1 ) 0.1 0.2 0.3 0.4 TC PD f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 r (day –1 ) 0.1 0.2 0.3 0.4 f (day –1 ) 0.3 0.2 0.1 FIG A6. Model-predicted treatment outcomes of radiotherapy (RT) alone, RT after surgical resection (SR), and RT before surgical resection (RS) for tumors of 10 8 viable cancer cells pretreatment and increasing surviving fractions of effector cells at 2 Gy (SF 2 ) in a cohort of virtual patients with different combinations of tumor growth rate (r) and immune recruitment rate (f) in response to tumor burden. Radiation is delivered to a total dose of 50 Gy in 25 daily fractions at 2 Gy per day, 5 days per week. Tumor control (TC) by treatment (blue) and progressive disease (PD; red) refer to tumor eradication and escape after treatment, respectively. For all simulations, we set RT-induced antitumor immunity to q= 4.5 ×10 −1 day −1 . L´ opez Alfonso et al 16 © 2019 by American Society of Clinical Oncology Downloaded from ascopubs.org by H. Lee Moffitt Cancer Center and Research Institute on April 18, 2019 from 206.081.166.225 Copyright © 2019 American Society of Clinical Oncology. All rights reserved.