Camargo Cancer Center Research Articles

Comparison of supervised machine learning classification techniques in prediction of locoregional recurrences in early oral tongue cancer

BackgroundThe proper estimate of the risk of recurrences in early-stage oral tongue squamous cell carcinoma (OTSCC) is mandatory for individual treatment-decision making. However, this remains a challenge even for experienced multidisciplinary centers. ObjectivesWe compared the performance of four machine learning (ML) algorithms for predicting the risk of locoregional recurrences in patients with OTSCC. These algorithms were Support Vector Machine (SVM), Naive Bayes (NB), Boosted Decision Tree (BDT), and Decision Forest (DF). Materials and methodsThe study cohort comprised 311 cases from the five University Hospitals in Finland and A.C. Camargo Cancer Center, São Paulo, Brazil. For comparison of the algorithms, we used the harmonic mean of precision and recall called F1 score, specificity, and accuracy values. These algorithms and their corresponding permutation feature importance (PFI) with the input parameters were externally tested on 59 new cases. Furthermore, we compared the performance of the algorithm that showed the highest prediction accuracy with the prognostic significance of depth of invasion (DOI). ResultsThe results showed that the average specificity of all the algorithms was 71% . The SVM showed an accuracy of 68% and F1 score of 0.63, NB an accuracy of 70% and F1 score of 0.64, BDT an accuracy of 81% and F1 score of 0.78, and DF an accuracy of 78% and F1 score of 0.70. Additionally, these algorithms outperformed the DOI-based approach, which gave an accuracy of 63%. With PFI-analysis, there was no significant difference in the overall accuracies of three of the algorithms; PFI-BDT accuracy increased to 83.1%, PFI-DF increased to 80%, PFI-SVM decreased to 64.4%, while PFI-NB accuracy increased significantly to 81.4%. ConclusionsOur findings show that the best classification accuracy was achieved with the boosted decision tree algorithm. Additionally, these algorithms outperformed the DOI-based approach. Furthermore, with few parameters identified in the PFI analysis, ML technique still showed the ability to predict locoregional recurrence. The application of boosted decision tree machine learning algorithm can stratify OTSCC patients and thus aid in their individual treatment planning.

Retroauricular endoscope-assisted versus conventional submandibular gland excision for benign and malignant tumors.

Surgical scars are a significant cosmetic problem, especially when in exposed areas such as the anterior neck. To avoid or reduce visible scarring, diverse innovative surgical approaches to the neck have been introduced. The purpose of this study was to evaluate the feasibility and safety of the endoscopic resection using the retroauricular approach for submandibular gland excision. The present study enrolled 48 patients who underwent conventional transcervical submandibular gland excision and 23 patients who underwent endoscope-assisted retroauricular approach submandibular gland excision, from February 2014 through February 2018 at the Department of Head and Neck Surgery and Otorhinolaryngology of the AC Camargo Cancer Center, in Sao Paulo, Brazil. The surgical outcomes were retrospectively reviewed. In the conventional group, 26 (54%) patients were male. The mean age was 49.3years (range 22-81). Twenty-two patients (46%) had sialoadenitis; twenty-one (44%) had benign and five (10%) had malignant tumors. The mean total surgical time was 86.4min (range 40-180), and the mean total length of hospital stay was 1.3days. Twenty-seven (56%) patients suffered from local postoperative complications in the neck. In the retroauricular group, 14 (61%) patients were male. The mean age was 44.1years (range 24-71years). Seven patients (31%) had sialoadenitis, twelve (53%) had benign tumors and four (16%) had malignant tumors. The mean total surgical time was 86.4min (range 75-300min), and the mean total length of hospital stay was 1.2days. Twelve (53%) patients suffered from local postoperative complications in the neck. No surgical site infections or systemic complications were described. The retroauricular endoscopic-assisted submandibular gland resection is feasible, with excellent cosmetic results and no significant complication rate increase, and can be a safe potential surgical alternative for patients who are motivated to avoid a visible neck scar.

Real Life CNS-IPI Score Validation: A Retrospective Analysis of Two Centers in Brazil Outside of Clinical Trial

Introduction:Diffuse large B cell lymphoma (DLBCL) is the most common type of Non-Hodgkin lymphoma (NHL). Treatment has improved over the years and can achieve 70% of survival (OS) in five years. Despite all progress in this field, central nervous system infiltration (CNSi), which is a devastating complication, still occurs with an incidence of 2-10%. Therefore, defining high-risk patients who consequently need prophylaxis seems to be the most adequate approach. Objective: Evaluate central nervous system international prognostic index (CNS-IPI) score regarding reproducibility and validity of the model in a heterogeneous cohort outside of clinical trial. Patients and methods: This retrospective research study was approved by the Santa Casa de Misericórdia de São Paulo Medical School (SCMSP) and Ac Camargo Cancer Center (ACCC) Ethics Committee. Patients were eligible if they were sequentially diagnosed with primary DLBCL, between January 2007 and January 2016. Also, if patients had complete chart data, pathological review and had started treatment at the respective institution (SCMSP or ACCC). Lymphoma staging was defined according tothe Ann Arbor system and patients were also categorized by the international prognostic index - IPI at baseline All patients were treated with rituximab-based chemotherapy, mostly R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone).CNSi was diagnosed by liquor (positive cytology and/or immunophenotype), computerized tomography, magnetic resonance image and/or fluorodeoxy-glucose-positron emission tomography, requested only in symptomatic patients when CNSi was clinically suspected. Confirmed cases were classified as parenchymal, leptomeningeal or both. Validation of the CNS‐IPI was assessed by graphical comparison between our results of Kaplan‐Meier curves (risk stratification groups: low, intermediate and high risk) and those previously defined by CNS‐IPI. Results:After inclusion and exclusion criteria 322 patients were available for the analysis (55% at ACCC vs. 46% at the SCMSP). Median follow-up was 60 months and median age was 58 years. There was similar distribution between male and female (47.8 vs. 52.2, respectively), and also similar distribution between localized (stage I-II) and advanced disease (stage III-IV), 47.2 vs. 51.2%, respectively. More than half of patients were low or low-intermediate IPI (59.6%) and low and intermediate CNS-IPI 0-3 (77.6%). The minority was submitted to CNS prophylaxis (19.9%), and the majority achieved complete response at the end of therapy (73.3%). Seven patients experienced CNSi, characterizing an incidence of 2.17% (7/322). The median time from diagnosis of DLBCL to CNSi was six months (range 2-52 months), while the median time of CNSi to death was 19 days (range 0 days -9.3 months). In our cohort, given the small number of events (7/322), an univariate model for risk factors was not possible, even though we statistically compared the groups of patients without and with CNSi. In univariate analysis at baseline LDH, number of extranodal sites, IPI, kidney/ adrenal were statically different in patients without and with CNSi (Table 1). Furthermore, regarding treatment, absence of complete response was also statistically different. CNS-IPI model stratified patients in a three-risk group model (low [0-1 points], intermediate [2-3 points] and high risk [4-6 points]) and demonstrateda 2-year rate of CNS relapse of 0.6%, 3.4% and 10.2%, respectively. In our cohort, using the same stratification, we obtained the2-year rate of CNS relapse of 0.0%, 0.8% and 13.8%, respectively. There was a statistical survival difference between the cohort of patients without or with CNSi. Median OS for patients without CNSi was not reached (average: 9.0 years [95% CI:8.5-9.5]) while median OS for patient with CNSi was of 8 months (95% CI: 2.8- 13.1). Log-rank p=0.000. Conclusion:CNSi in patients with DLBCL treated with rituximab based chemotherapy remains a rare but devastating complication. Our study reinforces the validity and reproducibility of CNS-IPI, specific outside of clinical trials. Also guarantee the safety of CNS-IPI use in daily practice. Our results demonstrate that almost 80% of patients are low or intermediate risk of CNSi with no need of interventions. Those considered high risk would benefit from more intensive diagnosis and prophylaxis approach. Table 1 Disclosures No relevant conflicts of interest to declare.

Complete metastasectomy in renal cell carcinoma: a propensity-score matched by the International Metastatic RCC Database Consortium prognostic model.

IntroductionWe evaluated overall survival (OS) benefit of complete metastasectomy (CM) in metastatic renal cell carcinoma (mRCC) using a propensity score-matched (PSM) analysis to balance groups by age, gender and by the International Metastatic RCC Database Consortium prognostic model (IMDC).MethodsWe included patients (pts) treated at the AC Camargo Cancer Center between 2007 and 2016. Pairs were matched by age, gender and IMDC. Kaplan–Meier survival estimates and Cox proportional hazard models were used to evaluate OS on CM and no-CM group.ResultsWe found 116 pts with clear cell mRCC. After PSM, the number was reduced to 74 (37 CM, 37 no-CM). The median OS for CM and no-CM was 98.3 months and 40.5 months, respectively (hazard ratio 0.24 95%CI 0.11–0.53 p < 0.001). The OS benefit of CM was confirmed on favourable and intermediate IMDC but was absent on poor IMDC. The CM group received less systemic therapy than the no-CM group. Ten pts in the CM group still have no evidence of disease (NED).ConclusionAfter matching for age, gender and IMDC, we found CM impacts on OS and also diminishes the need for systemic treatment. Survival benefit was confirmed for favourable/intermediate IMDC but not for the poor IMDC prognostic model. Further studies correlating IMDC and metastasectomy are needed to guide clinical decision-making.

Assessment of Tumor-infiltrating Lymphocytes Predicts the Behavior of Early-stage Oral Tongue Cancer.

Tumor-infiltrating lymphocytes (TILs) have shown a promising prognostic value in many epithelial cancers. We sought to assess the prognostic value of TILs in a multicenter cohort of early oral tongue squamous cell carcinoma (OTSCC). The percentage of TILs was assessed on the surgical resection slides stained with hematoxylin and eosin. The assessment of TILs was performed in the stromal compartment and in the intraepithelial compartment (at the invasive front and at the center of the tumor). We followed the method that was described recently by the International Immuno-Oncology Biomarker Working Group for the assessment of TILs. A total of 308 cases from the 5 Finnish university hospitals and from A.C. Camargo Cancer Center, São Paulo, Brazil, were included. We found a promising prognostic value for stromal TILs at the invasive front in the multivariable analysis with a hazard ratio of 2.61 (95% confidence interval [CI], 1.77-3.83; P<0.001) for overall survival, 1.99 (95% CI, 1.07-3.69; P=0.040) for disease-specific survival, and 1.94 (95% CI, 1.14-3.29; P=0.020) for disease-free survival. In conclusion, evaluation of TILs is simple and can aid in identifying the high-risk cases of early OTSCC. The method introduced by the International Immuno-Oncology Biomarker Working Group can be used for standardized determination of TILs in early OTSCC.

1404TiP - HORMONET: Study of tamoxifen in well differentiated neuroendocrine tumours and hormone receptor positive expression

BackgroundPatients (pts) with advanced neuroendocrine tumors (NET) represent an unmet medical need, what motivates the development of new treatments. Prior retrospective study conducted by our group showed that among 96 NET, immunohistochemistry (IHC) expression for estrogen (ER) and progesterone receptors (PR) were observed in 20.8% and 18.8% of cases, respectively. Additionally, an old clinical trial (Mortel C, 1984) suggested antitumor effects of estrogen/progesterone-directed therapies in NET pts. Yet the effects of an antiestrogen agent in NET pts whose tumor express ER and/or PR are unknown. Hence we will conduct a clinical trial to test the efficacy of tamoxifen in ER/PR positive NET pts. Trial designSingle-arm phase II multicenter trial of tamoxifen 20mg PO continuously until intolerance and/or tumor progression. Eligible pts must be≥18 years old, have histologically confirmed advanced, irresectable well-differentiated lung or gastroeteropancreatic NET, IHC expression of ER and/or PR≥1%, documented progression in the prior 12 months, measurable disease, prior treatments with standard therapies and/or no indication to start new treatment due to lack of access, risk of toxicities or without clinical indication (patients who meet criteria for watchful waiting may be included), ECOG 0-2, adequate organ function. Main exclusion criteria are: aggressive disease requiring cytotoxic therapy, use of oral anticoagulants, previous history of deep vein thrombosis or pulmonary embolism in the last 12 months, postmenopausal vaginal bleeding with no defined etiology. The primary endpoint is disease control rate (DCR) at week 24 as per RECIST 1.1. Secondary endpoints are progression-free survival, biochemical response, response rate. Exploratory evaluations: circulating tumor cells kinetics (qualitative and quantitative) and PET-CT gallium-68 intake intensity variation (at baseline and at week 24). Sample size assumptions: the H0 is DCR at week 24 of 50% and H1, 70%; with a type I error of 10%, power of 80% and attrition rate of 30%, the final sample size will be 22 pts. Clinical trial identificationNCT03870399. Legal entity responsible for the studyRachel Riechelmann. FundingAC Camargo Cancer Center. DisclosureM. Barros: Honoraria (self): Novartis. J.F. Rego: Honoraria (self): Novartis. R.P. Riechelmann: Honoraria (self): Novartis. All other authors have declared no conflicts of interest.

660TiP - BISQUIT: A randomized phase II study of the administration of prebiotics and probiotics during definitive treatment with chemotherapy-radiotherapy for patients with squamous cell carcinoma of the anal canal

BackgroundThe standard treatment for patients (pts) with localized squamous cell carcinoma of the anal canal (SCCAC) is chemoradiation (ChRT). Yet nearly 30% of patients are not cured and undergo salvage anorectal amputation. SCCAC is mostly a virus-associated tumor and thus potentially immunogenic. In fact, immune checkpoint inhibitors are promising in trials of metastatic SCCAC. Recently, studies have shown that the composition of the intestinal microbiota influences response to immunotherapy in some solid tumors, and the replacement of the intestinal "carcinogenic" by a protective microbiota has led to investigations with prebiotics and probiotics (PreProbiotics). Yet, there are no studies on the use of these agents in SCCAC. Thus we are conducting a randomized phase II study to test the efficacy of PreProbiotics during definitive ChRT aiming to improve the cure rate of pts with localized SCCAC. Trial designRandomized open label parallel phase II trial, where eligible pts will be randomized 1:1 to receive PreProbiotics starting one week prior to ChRT and throughout treatment until response evaluation at 6 to 8 weeks post ChRT or conventional ChRT. Eligible pts are ≥ 18 years, with histologically confirmed SCCAC, localized disease (≥ T2N0M0), indication to start definitive ChRT; HIV seropositive pts are eligible. Pts with active infection requiring antibiotics will be excluded. Primary endpoint: complete clinical and radiological response (CR) at 6 to 8 weeks post ChRT. Secondary endpoints: CR at 6 months, progression free survival, colostomy-free survival, metabolic response measured by 18-FDG PET-CT (baseline and at 6-8 weeks), toxicity, incidence of HPV in tumor tissues. All pts will have the following biological samples collected for correlative studies at baseline, 6-8 weeks and 6 months post ChRT: blood samples for circulating tumor DNA, inflammatory cytokines, anal/rectal swabs and feces to evaluate microbiota. Sample size assumptions: the H0 is CR at week 6 - 8 of 70% and H1, 90%; with a type I error of 10%, power of 80% and attrition rate of 10%, the final sample size is 75 patients. Clinical trial identificationNCT03870607. Legal entity responsible for the studyRachel Riechelmann. FundingAC Camargo Cancer Center. DisclosureAll authors have declared no conflicts of interest.

A modified recursive partitioning analysis for predicting overall survival in patients with non-small cell lung cancer and central nervous system metastases.

Non-small cell lung cancer (NSCLC) is a major cause of brain metastases. Nonetheless, patients with central nervous system (CNS) spread are poorly represented in clinical trials. We sought to evaluate the overall survival (OS) of patients with NSCLC and CNS metastases. Patients with NSCLC and CNS metastases treated at A. C. Camargo Cancer Center from January 2007 to December 2017 were selected. The primary endpoint was OS following the diagnosis of CNS metastasis. The Kaplan-Meier method was applied to calculate OS. Prognostic factors were assessed by the Cox Proportional Hazards model. As an exploratory analysis, a survival tree was generated based upon the two most statistically significant variables in the multivariate model and one additional clinically meaningful variable. In total, 311 patients were included. Median OS was 10.3 months (95% CI, 8.7-13.1 months). ECOG performance status 2-4 (HR 2.12; 95% CI, 1.40-3.20; P<0.01) and the absence of a driver mutation (HR 3.30; 95% CI, 1.85-5.90; P<0.01) were strongly associated with worse OS. A Modified Recursive Partitioning Analysis (mRPA) was developed based on the curves generated by the survival tree. mRPA stratified our cohort in four subgroups with significantly different OS (3.1 to 43 months) and it outperformed both RPA and GPA in predicting OS in our population. OS in our cohort was better than previously reported. However, prognosis is widely variable and is mostly dictated by performance status and the presence of a driver mutation.

Comparison of enzalutamide versus abiraterone in castration-resistant prostate cancer before docetaxel: Results of a propensity score-matched analysis.

e16540 Background: Metastatic castration-resistant prostate cancer (mCRPC) phenotype involves androgen-receptor signalling mechanisms that support the use of enzalutamide (EZ) and abiraterone (Abi). These therapies improve overall survival (OS) and quality-of-life, with a favourable safety profile. There is no validated data defining the best drug or sequence to be used. Methods: A retrospective cohort of mCRPC patients (pts) was analysed at AC Camargo Cancer Center. The primary objective was to compare progression-free survival (PFS) and OS in patients treated with EZ or Abi as first line (docetaxel naïve pts). Kaplan-Meier, Log-Rank Test and Cox Regression were used for survival analysis. To address unbalanced characteristics between the two treatment groups treatment efficacy was compared in a propensity score matched cohort. Results: From May, 2002 to September, 2017, 120 pts were treated with Abi (84%) or EZ (36%). Median follow-up was 21.2 months. Median age at diagnosis was 66 (48-84), the majority were Gleason score 7 (34%) and median baseline PSA was 14. Median PFS was 17.4 months in EZ and 10.6 months in Abi group (HR = 0.65; 95%CI: 0.39-1.10; p = 0.11). Median OS was not reached and 31.6 months for EZ and Abi, respectively (HR = 0.60; 95%CI: 0.27-1.36; p = 0.22). EZ was associated with a better PSA response in the first 4 months of treatment (p &lt; 0.001). Independent prognostic factors for worse OS and PFS were ECOG ≥ 1, treatment toxicity ≥ G1 and lower PSA doubling time before treatment and for better OS and PFS were PSA response in the first 4 months and alkaline phosphatase and lactate dehydrogenase response at any time. In the propensity score matched cohort including 72 patients PFS was better in EZ group (HR = 0.36; 95%CI: 0.20-0.64; p &lt; 0.001) but there was no difference in OS (HR = 0.66; 95%CI: 0.27-1.63; p = 0.37). Conclusions: EZ was associated with prolonged PFS and better PSA response, with no OS improvement when compared with Abi for mCRPC before docetaxel, in a propensity score matched analysis.

Disassociation of ARID1A with genomic ancestry and prognostic impact in an admixed cohort of Brazilian patients with gastric cancer (GC).

e15585 Background: Loss of expression of ARID1A, a tumor suppressor gene involved in chromatin remodeling and transcription activation, has been associated with worse prognosis in Asian GC patients (Yang et al. 2016). Mutations in ARID1A have been found in 8-27% of GC, usually leading to gene/protein inactivation. Here we evaluated the possible involvement of ARID1A mutations and clinical characteristics, while considering genomic ancestry and survival, in a cohort of Brazilian GC patients. Methods: We included 112 pts diagnosed with GC and treated at AC Camargo Cancer Center before 2013. The study was approved by local IRB. Genomic DNA was used for capture-based enrichment of a customized gene panel including 99 genes. Libraries were sequenced in the NextSeq 500 platform (Illumina), using paired-end reads (2x75bp). For ancestry inference we used a set of ancestry informative markers, covered by target and off-target reads, described by Elhaik et al. (2014). Results: Median age was 64y (37-91), 63% were male, M:F ratio was 1.73. Most cases were classified as Diffuse (47.3%) followed by Intestinal (41.1%), Mixed (2.7%) and 8.9% were deemed unclassifiable by Lauren´s classification. 22.3% were stage I, 20.5% stage II, 42.9% stage III and 14.3% stage IV. 11.6% were in the GEJ and 80.4% were in corpus/antrum. Five patients were EBV positive (4.5%). Genomic ancestry was as follows: 55.4% European, 27.7% Asian, 8.9% African and 8% were highly admixed ( &lt; 50% of any ancestry). ARID1A was mutated in 19% of cases and showed no association with age at diagnosis (p = 0.21), gender (p = 0.76), tumor location (p = 0.55), staging (p = 0.42), Lauren (p = 0.14) or EBV (p = 0.42). ARID1A had no impact on overall survival (OS) (HR 1.17; 95%CI 0.59-2.31; p = 0.6) or disease-free survival (DFS) (HR 1.24; 95%CI 0.66-2.32; p = 0.5), including the subgroup with Asian genomic background: OS-Asia (HR 1.08; 95%CI 0.31-3.81; p = 0.9), DFS-Asia (HR 1.15; 95%CI 0.32-4.12; p = 0.8). Conclusions: ARID1A is a common driver in GC among Brazilian patients. Unlike in Asians, ARID1A was not prognostic in this Brazilian cohort even in the subgroup with a predominant ( &gt; 50%) Asian genomic ancestry.

Survival of gastric cancer (GC) patients is not determined by the predominant genomic ancestry (PGA): Results from an ethnically admixed Brazilian cohort of GC patients.

e15588 Background: Clinical characteristics, treatment response and overall survival of GC patients differ between Asian and non-Asian countries. Here we evaluated the possible associations between PGA, clinical characteristics and survival in an admixed GC Brazilian cohort. Methods: We included 112 GC pts diagnosed and treated at AC Camargo Cancer Center (São Paulo, Brazil) before 2013. The study was approved by local IRB. Genomic DNA was used for capture-based enrichment of a customized gene panel including 99 genes. Libraries were sequenced in the NextSeq 500 platform (Illumina), using paired-end reads (2x75bp). Ancestries were determined through a set of ancestry informative markers (AIMs), covered by target and off-target reads, described by Elhaik et al. (2014). Results: An average of 406 AIMs were recovered from the available samples, revealing average ancestries was as follows: 55.4% European, 27.7% Asian, 8.9% African; 8% of subjects were highly admixed (HA; &lt; 50% of any ancestry). We found no association between PGA and age at diagnosis (p = 0.58), tumor location (p = 0.34), Lauren (p = 0.24) and staging (p = 0.68). There was an association between PGA and gender (p = 0.04) and a marginal association between PGA and EBV (p = 0.056). BRCA2 was the only gene enriched in the Asian subgroup, compared to the other groups combined (p = 0.009). The median follow-up time was 95 months. We found no differences in median overall survival (OS) (p = 0.4) or disease-free survival (DFS) (p = 0.6) according to PGA. The HA subgroup presented worst survival outcomes compared to the other groups aggregated (mOS 34m; 95%CI 5-80 x mOS 71m; 95% CI 44-85, respectively), but the difference was not statistically significant (HR 1,73; 95% CI 0.74-4.05; p = 0.2). Even for patients with &gt; 75% AIMs for any given ancestry, we found no differences in OS between Europeans (n = 25; mOS 86m; 95%CI 80-NA), Asians (n = 22; mOS 83m; 95%CI 51-NA) and Africans (n = 5; mOS 67m; 95%CI 22-NA), p = 0.4. Conclusions: The most prevalent ancestries in this Brazilian GC cohort were European, followed by Asian and African. Although we found associations between ancestry and a few clinical aspects, PGA was not associated with survival.

A modified recursive partitioning analysis for predicting overall survival in patients with non-small cell lung cancer and central nervous system metastases.

e20604 Background: Non-small cell lung cancer (NSCLC) accounts for 85% of lung cancers. Most patients with NSCLC have metastases at diagnosis and 30-50% of them will present brain lesions during follow-up. Nonetheless, patients with central nervous system (CNS) spread are poorly represented in clinical trials and the management of this clinical scenario is not standardized. Prognostic tolls are needed to help to guide the best approach for these patients. Methods: Descriptive, analytical, retrospective, single center study. Patients &gt; 18 years of age diagnosed with NSCLC who developed CNS metastases treated at AC Camargo Cancer Center from January 2007 to December 2017 were included. The primary endpoint was overall survival (OS). The secondary endpoint was intracranial progression-free survival (IC-PFS). An exploratory analysis of prognostic factors associated with OS was undertaken. The Kaplan-Meier method was used to calculate OS and IC-PFS. We used the Cox Proportional Hazard model to assess the prognostic role of clinical and pathological characteristics on IC-PFS and OS. Variables with p &lt; 0.20 at univariate analysis were used to generate a multivariate model. A survival tree was generated based on the two most statistically significant variables in the multivariate model. A p &lt; 0.05 was considered statistically significant. Results: 311 patients were included. Median age was 60 years-old (IQR 54-68), 72.6% had an ECOG performance status 0 or 1 and 18.3% had a driver mutation detected ( EGFR, ALK or ROS1). IC-PFS was 7.1 months (95%CI 6.1-8.6) and OS was 10.3 months (95%CI 8.7-13.1). At multivariate analysis, the two most statistically significant prognostic factors associated with poor OS were ECOG performance status 2-4 (HR 2.12; 95%CI 1.40-3.20; p &lt; 0.01) and the absence of a known driver mutation (HR 3.30; 95%CI 1.50-3.87; p &lt; 0.01). Based on the curves generated by the survival tree, a prognostic model was developed - the Modified Recursive Partitioning Analysis (mRPA). This model stratified our cohort in four subgroups with significantly different OS (3.1 to 43 months). mRPA was better than RPA and GPA to predict prognosis in our population. Conclusions: OS and IC-PFS in patients with NSCLC and CNS metastases were better than previously reported. However, prognosis is widely variable and is directed mainly by performance status and the presence of a driver mutation. We propose a new prognostic model that my help to guide the treatment of these patients in the future. This model needs external validation.

Preoperative vitamin D deficiency is a risk factor for postoperative hypocalcemia in patients undergoing total thyroidectomy: retrospective cohort study.

The relationship between preoperative vitamin D deficiency and postoperative hypocalcemia in cases of total thyroidectomy (TT) is a matter of controversy and may vary according to geographical scenarios and populations. The objective here was to evaluate whether preoperative vitamin D deficiency was associated with postoperative symptomatic hypocalcemia in a population in South America. Retrospective cohort study on data from all patients undergoing total thyroidectomy, with or without central compartment lymph node dissection, from January 2014 to December 2017, at the A. C. Camargo Cancer Center. Patients with benign thyroid disease (Graves' disease, multinodular goiter or hyperthyroidism) or thyroid cancer who underwent primary total thyroidectomy with or without central compartment lymph node dissection were included. The exclusion criteria were simultaneous parathyroidectomy and conditions that could affect serum calcium levels. The data collected included patient demographics, thyroid pathology, extent of the surgical procedure and complications. Information on preoperative and postoperative calcium, parathyroid hormone (PTH) and vitamin D levels were retrieved from the medical records. 1,347 patients were assessed and postoperative hypocalcemia was diagnosed in 284 patients (21%). The vitamin D levels were considered deficient in 243 patients (18%). Postoperative hypocalcemia was diagnosed in 357 patients (31.5%). Multivariate analysis showed that central compartment dissection and preoperative total calcium and deficient vitamin D levels were significant risk factors for postoperative hypocalcemia. Deficient preoperative vitamin D levels were a significant risk factor for postoperative hypocalcemia. Preoperative oral supplementation should be considered, to minimize this risk.

Defective regression models for cure rate modeling with interval‐censored data

Regression models in survival analysis are most commonly applied for right-censored survival data. In some situations, the time to the event is not exactly observed, although it is known that the event occurred between two observed times. In practice, the moment of observation is frequently taken as the event occurrence time, and the interval-censored mechanism is ignored. We present a cure rate defective model for interval-censored event-time data. The defective distribution is characterized by a density function whose integration assumes a value less than one when the parameter domain differs from the usual domain. We use the Gompertz and inverse Gaussian defective distributions to model data containing cured elements and estimate parameters using the maximum likelihood estimation procedure. We evaluate the performance of the proposed models using Monte Carlo simulation studies. Practical relevance of the models is illustrated by applying datasets on ovarian cancer recurrence and oral lesions in children after liver transplantation, both of which were derived from studies performed at A.C. Camargo Cancer Center in São Paulo,Brazil.

Abstract P1-03-10: The implementation of a patient referral service in a Brazilian cancer center

Abstract BACKGROUND: Healthcare systems face problems of cost increases and poor delivery organization. Integrated delivery may reduce costs and improve quality and health outcomes. OBJECTIVE: To describe how the A.C. Camargo Cancer Center, aiming at greater integration, implemented a referral service for breast cancer patients. METHODS AND RESULTS: The process was divided in three phases: 1) AS IS ANALYSIS, 2) BENCHMARKING, 3) PILOT &amp; IMPLEMENTATION. 1) AS IS ANALYSIS A.C. Camargo was responsible for treating 16% of all breast cancer cases that arose from 2000 to 2012 in São Paulo State. Prior to implementation of the program, there was no special patient classification at the initial appointments. 2) BENCHMARKING As proposed by MD Anderson Cancer Center, we used the patient's classification at the time of an appointment request (regular screenings, undiagnosed or breast cancer patient). As proposed by Memorial Sloan Kettering Cancer Center, we used a "Physician Referral Service" staffed by "Referral Specialists" and "Trained Oncology Nurses" to collect patient information prior to the first appointment. As suggested by Princess Margaret Cancer Center, we collected patient care data (e.g., abnormal imaging, palpable lump). 3) PILOT &amp; IMPLEMENTATION Phase I: Feb. 13, 2017 to Dec. 28, 2017. We reached 7% of new patients in the Breast Surgery Department (BSD): 48% were in the Cancer Group, 45% in the Abnormal Imaging Group (Undiagnosed) and 7% in the Palpable Lump Group. New patients were classified by the Call Center. Electable cases were referred to the Nurse Navigator, who proceeded with appointments according to protocol. Root cause analysis of non-captured patients led to the following improvements: extension of participant Call Center cells, script review, and implementation of a training program. The Phase I results led to the following improvements: 1) reclassification of a subgroup with highly suspicious images as the "Cancer Group"; 2) Transfer of the new referral and scheduling functions to the Call Center; and 3) Implementation of new "first appointment items", personalized for each patient group. This information was displayed to the physician and operations staff in advance of the consultation. Phase II: Dec. 29,2017 to June 28, 2018. We reached 100% of new patients in the BSD: 17% were in the Cancer Group, 23% in the Abnormal Imaging Group, 8% in the Palpable Lump Group, and 53% in the Regular Screening Group. Phase II results led to the following improvement: 1.) Distribution of the cancer groups'first appointment items equally among surgeons. Phase III: From June 29, 2018 to the operation. Implemented new appointment items for new clinical patients and for pre-treatment returns. CONCLUSION: With implementation of the Referral Service, the BSD at A.C. Camargo is now able to identify the reason for each appointment before the first consultation. This practice promotes operational predictability and more effectively organizes the personalized journey of each patient, including care on the part of Nurse Navigators for the most critical cases. REFERENCE: Institute of Medicine (US) Committee on Quality of Health Care in America. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington (DC): National Academies Press (US); 2001 Citation Format: Makdissi FB, Costa Filho ER, Conti EC, Santos LC. The implementation of a patient referral service in a Brazilian cancer center [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P1-03-10.

Abstract P1-19-08: Prognostic factors of human epidermal growth factor 2 (HER2) positive breast cancer patients with brain metastasis in the context of all available new therapies

Abstract Background: Patients with advanced human epidermal growth factor 2 (HER2)-positive breast cancer are at the highest risk developing breast cancer brain metastases (BCBM), which are associated with significant morbidity and mortality. The advent of HER2-directed therapy resulted in greatly improved survival outcomes and have been widely used. In this context of a better controlled systemic disease, brain metastases are emerging as a new challenge for the oncologist. The aim of this study was to determine the clinicopathologic factors associated with the prognosis of patients with HER2 positive BCBM, in the scenario of new therapies. Methods: A retrospective review of clinical data from patients who have developed HER2 positive BCBM from March 2005 to January 2018 at AC Camargo Cancer Center. Patients characteristics were collected, and overall survival (OS) , time to first HER2 positive BCBM and OS after HER2 positive BCBM were estimated by the Kaplan-Meier method. Associations between OS after BCBM and clinical variables were assessed by Cox proportional hazards regression models. Results Among 80 patients with HER2-positiveBCBM median age was 51 years (range, 24–81 years). Of the patients, 55 patients (69%) had ER-positive/HER2-positive breast cancer, and 25 (31%) had ERnegative/ HER2-positive breast cancer. The median brain metastasis-free survival period from primary breast cancer was 33.5 months .The median survival after developing brain metastasis was 28 .5 months. Patients with more than 3 brain metastases had significantly shorter overall survival(p=0,01). ). Nineteen patients (24%) underwent surgical treatment of metastasis in the brain and there was no associated with survival duration(p= 0,33). Patients treated with radiotherapy 71 (88%) had significantly better survival (p= 0,013), if the patient received stereotactic radiosurgery (53.5%) compared to total brain radiotherapy received ( 46,5%) was not associated with duration of survival (p= 0,24). . Treatment with other HER2-targeting agents after BCMB had significantly increased survival (p=0,022), including trastuzumab 53 (67%), pertuzumab 21 (26%), trastuzumab emtansine (TDM1 )= 35 (43%) and lapatinib 37 ( 46%). . In multivariate analysis, the presence of more than 3 brain metastases (p=0,005) was the sole independent prognostic factors. Conclusions: Our study indicates that HER2-positive patients with more than three brain metastases had a poor prognosis and that regardless of the use of new therapies. The role of new therapies in the management of patients with brain metastases is often not clearly defined. The best treatment strategy is not yet known and the study of prognostic factors may help to choose a better treatment sequence. Citation Format: Medeiros de Melo JC, da Costa AA, de Castro DG, Gomes AA, Guimaraes AP, Sanches SM, Diniz AC, Ribeiro AR. Prognostic factors of human epidermal growth factor 2 (HER2) positive breast cancer patients with brain metastasis in the context of all available new therapies [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P1-19-08.

Characteristics and survival of older patients with metastatic pancreatic cancer: a retrospective analysis of the AC Camargo Cancer Center experience.

Background:Advanced age is the most important risk factor for pancreatic cancer and about half of patients are diagnosed with metastatic disease. In the first-line setting, multidrug chemotherapy regimens were shown to be more effective than gemcitabine alone. However, the older population was under-represented in randomized clinical trials. We aimed to describe the clinical profile of older patients with metastatic pancreatic cancer and their survival outcomes.Materials and methods:This was a retrospective, unicentric study that included patients diagnosed with metastatic pancreatic cancer (non-neuroendocrine), aged 65 years and over.Results:The study population comprised 196 patients. The median age was 73 years; 67% of these patients presented Eastern Cooperative Oncology Group performance status (ECOG) ⩽ 1 and the median Charlson Comorbidity score was 10. Chemotherapy was given to 89% of the patients. The most frequently used chemotherapy regimens were gemcitabine (44%), 5-fluorouracil and oxaliplatin [FOLFOX; 26%], and 5-fluorouracil, oxaliplatin and irinotecan (FOLFIRINOX; 20%). Patients treated with FOLFIRINOX were younger and they presented better performance status. After a median follow up of 19.8 months, the median overall survival (OS) was of 7.2 months and the median time to first-line-treatment failure was 4.6 months. Among patients treated with chemotherapy, the median OS was highest for those treated with FOLFIRINOX (13.8 months), as compared with FOLFOX (7.0 months) or gemcitabine (6.7 months); p = 0.004. Nonetheless, treatment with FOLFIRINOX was associated with increased risk of severe toxicity (p = 0.008).Conclusion:Older patients with metastatic pancreatic cancer benefit from palliative chemotherapy, and FOLFIRINOX is a therapeutic option in rigorously selected older patients.

Clinico-pathological features and survival of patients with malignant exocrine pancreatic neoplasms: The AC Camargo Cancer Center experience.

Pancreatic cancer plays an important role in cancer-related mortality. Few studies have been performed in Brazil to characterize patients affected by this disease. We aimed to describe the clinico-pathological characteristics and the survival of patients with pancreatic cancer seen at AC Camargo Cancer Center (ACCCC). We included patients ≥ 18-year old, with a histologically confirmed diagnosis of exocrine pancreatic cancer, that attended at least one visit at ACCCC from 2008 to 2016. The study included 739 patients. Median age at diagnosis was 64 years. Most patients were male. About 5% presented a family history of pancreatic cancer. A total of 40% had diabetes and 51.4% presented with ECOG performance status 1. Tumors most often arose in the pancreatic head and roughly half of the patients had metastatic disease at presentation. Median overall survival of patients with potentially resectable disease submitted to surgery at ACCCC was 35.4 months. Median overall survival times of patients with the unresectable and metastatic disease were 14.1 and 9.3 months, respectively. The features of our population match those of studies done in developed countries. We believe multicentric data from patients with pancreatic cancer in Brazil could enable more effective preventive and therapeutic approaches to the disease.

1742P - Geriatric oncology (GO): Long term follow-up of a prospective series of the GO unit at A.C. Camargo Cancer Center

Background: Ageing is a global phenomenon and has huge impact in healthcare. Still emerging in developing nations, there has been no major national innitiative focusing on oncologic geriatric patients (pts). The study describes a program to > 70y cancer pts in A C Camargo Cancer Center. We hypothesized that incorporating CGA into daily practice can improve individualized care. Methods: Previous aims were to evaluate if: 1) CGA could be feasible in daily practice; 2) it could be useful as a tool in treatment/tx decision making; 3) analyse the impact of CGA in the selection of tx, with its ability to predict complications (such as dosis reduction, hospitalization, or tx discontinuation) in an Oncogeriatric Unit. The present study is to present the long term survival of those pts, candidates to receive systemic tx, who underwent CGA assessments, which included scales of: ADL (Katz, Lawton), mini-nutritional assessment, depression (GDS), comorbidities and polypharmacy. Patients were classified as fit, borderline or frail. Fit pts received mainly full treatment; frail/borderline pts, mainly modified tx or specific supportive care. Results: From Oct 25/10-Dec/12/12, 638 pts were included to be prospectively evaluated - female 55%; cancers (%): breast 26, prostate 12, colorectal 12, hematologic 8. Katz A = 68%, Lawton 27=45%; polipharmacy ( > =5): 48%; depression 29%, undernutrition: 21%. We have previously presented (ASCO 2014) that there was an association between the choice of oncologic treatment (original vs modified or with additional dose reduction) with Katz, Lawton, depression and nutrition scales (p < =0.035); as well as the ability to complete the individualized proposed treatment, which was correlated with Lawton (p = 0.04) scale. Hospitalization during chemotherapy was associated with comorbidities, altered Katz, Lawton and depression scales (p < =0.038). With 610 pts evaluable, median survival was 51.7 months. Conclusions: We could find a high long term survival of 70+ cancer pts, showing CGA to be feasible and allowing long term survival follow-up. However, many pts have been lost to follow-up. How to better perform in pts' adherence and to reduce bias must be a concern in real world geriatric oncology. Legal entity responsible for the study: Aldo L.A. Dettino; Grupo de Assistencia, Ensino e Pesquisa em Oncologia (GAEPO). Funding: Has not received any funding. Disclosure: All authors have declared no conflicts of interest.

P1.15-07 Geriatric Oncology - Lung Cancer in Elderly Patients and Palliative Treatment: Prospective Analysis of A. C. Camargo Cancer Center

P1.15-07 Geriatric Oncology - Lung Cancer in Elderly Patients and Palliative Treatment: Prospective Analysis of A. C. Camargo Cancer Center