Development of an artificial intelligence-derived histologic signature associated with adjuvant gemcitabine treatment outcomes in pancreatic cancer

Pancreatic ductal adenocarcinoma (PDAC) has been left behind in the evolution of personalized medicine. Predictive markers of response to therapy are lacking in PDAC despite various histological and transcriptional classification schemes. We report an artificial intelligence (AI) approach to histologic feature examination that extracts a signature predictive of disease-specific survival (DSS) in patients with PDAC receiving adjuvant gemcitabine. We demonstrate that this AI-generated histologic signature is associated with outcomes following adjuvant gemcitabine, while three previously developed transcriptomic classification systems are not (n = 47). We externally validate this signature in an independent cohort of patients treated with adjuvant gemcitabine (n = 46). Finally, we demonstrate that the signature does not stratify survival outcomes in a third cohort of untreated patients (n = 161), suggesting that the signature is specifically predictive of treatment-related outcomes but is not generally prognostic. This imaging analysis pipeline has promise in the development of actionable markers in other clinical settings where few biomarkers currently exist.

Predictors of spine metastases at initial presentation of pediatric brain tumor patients: a single-institution study

PURPOSE

Given the rarity of disseminated disease at the time of initial evaluation for pediatric brain tumor patients, we sought to identify clinical and radiographic predictors of spinal metastasis (SM) at the time of presentation.

METHODS

We performed a single-institution retrospective chart review of pediatric brain tumor patients who first presented between 2004 and 2018. We extracted information regarding patient demographics, radiographic attributes, and presenting symptoms. Univariate and multivariate logistic regression was used to estimate the association between measured variables and SMs.

RESULTS

We identified 281 patients who met our inclusion criteria, of whom 19 had SM at initial presentation (6.8%). The most common symptoms at presentation were headache (n = 12; 63.2%), nausea/vomiting (n = 16; 84.2%), and gait abnormalities (n = 8; 41.2%). Multivariate models demonstrated that intraventricular and posterior fossa tumors were more frequently associated with SM (OR: 5.28, 95% CI: 1.79-15.59, p = 0.003), with 4th ventricular (OR: 7.42, 95% CI: 1.77-31.11, p = 0.006) and cerebellar parenchymal tumor location (OR: 4.79, 95% CI: 1.17-19.63, p = 0.030) carrying the highest risk for disseminated disease. In addition, evidence of intracranial leptomeningeal enhancement on magnetic resonance imaging (OR: 46.85, 95% CI: 12.31-178.28, p < 0.001) and hydrocephalus (OR: 3.19; 95% CI: 1.06-9.58; p = 0.038) were associated with SM.

CONCLUSIONS

Intraventricular tumors and the presence of intracranial leptomeningeal disease were most frequently associated with disseminated disease at presentation. These findings are consistent with current clinical expectations and offer empirical evidence that heightened suspicion for SM may be prospectively applied to certain subsets of pediatric brain tumor patients at the time of presentation.

Gemcitabine response prediction in the adjuvant treatment of resected pancreatic ductal adenocarcinoma using an AI histopathology platform

e16295

Background: Adjuvant chemotherapy improves survival following resection of pancreatic ductal adenocarcinoma (PDAC). A modified fluorouracil/irinotecan/oxaliplatin regimen (mFOLFIRINOX) has demonstrated improved disease free survival and overall survival, though gemcitabine-based monotherapy and gemcitabine plus capecitabine are alternatives in less fit patients. Though there are several proposed biomarkers to guide treatment decisions (GATA6, hENT1, and GemPred), no biomarker is used to guide treatment selection in clinical practice. Consequently, we sought to develop an artificial intelligence-derived signature of features from digital images of routine histopathology specimens that could identify patients susceptible to routine chemotherapeutic agents. Methods: 139 whole-slide digitized histological slides corresponding to 102 resected PDAC tumors from TCGA-PAAD were used in this study. This dataset corresponded to patients that had received either gemcitabine-backbone or 5 FU-backbone chemotherapy as their first-line adjuvant treatment. We extracted nuclei images from tissue regions using segmentation models and computed geometric features of these nuclei which we then correlated with Disease Specific Survival (DSS) in order to construct a signature associated with treatment benefit. This signature was compared against two board certified pathologists using the grade of the digital slides images to classify patients into above or below average DSS buckets. Results: Among quantitative geometric features, a set of area and ellipse features describing nuclei geometry correlated most with response to gemcitabine (R̃0.4). The cox proportional hazards model using these geometric nuclei features was found to be predictive of response to gemcitabine and achieved a C-index (95% CI) of 0.69 (0.58, 0.79). The pathologist-based baseline model for above and below average DSS had a median DSS of 443 and 461 days respectively. Using the average expected lifetime as the threshold, the model divides patients receiving gemcitabine into two histological subtypes with median DSS of 586 and 394 days respectively (p < 0.05). The model appeared specific to gemcitabine. Among patients receiving 5-FU (n = 10) there was no statistical significance in median DSS between the subtypes and a c-index of 0.63 (0.27, 1.0). Conclusions: An artificial intelligence approach utilizing only routine histopathology can identify features that correlate with treatment outcomes in PDAC with classification performance (c-index:0.69) superior to the validated AJCC treatment prediction tool (0.59).

Socioeconomic Disparities in Brain Metastasis Survival and Treatment: A Population-Based Study

OBJECTIVE

In the present study, we used a validated socioeconomic status (SES) index and population-based registry to identify and quantify the impact of SES on access to treatment and overall survival for patients diagnosed with synchronous brain metastases.

METHODS

The Surveillance, Epidemiology, and End Results database was used to extract all patients between 2010 and 2016 with brain metastases at initial presentation. SES was stratified into tertiles and quintiles using the validated Yost index. Multivariable logistic regressions were used to evaluate the impact of demographic, tumor, and socioeconomic covariates on receipt of radiotherapy and chemotherapy. Kaplan-Meier curves were used to estimate survival.

RESULTS

Between 2010 and 2016, 35,595 patients presented with brain metastases at the time of primary cancer diagnosis. Most patients received radiation and/or chemotherapy as part of the initial course of their treatment; 71.6% (n = 25,484) were irradiated while 54.4% (n = 19,371) received chemotherapy and 44.9% (n = 15,984) received chemoradiation. Patients in the highest Yost tertile and quintile experienced longer overall survival (P < 0.001). Additionally, multivariable logistic regression revealed that the lowest Yost quintile was significantly less likely to receive either radiation (adjusted OR: 0.82; 95% confidence interval: 0.75-0.89; P < 0.001) or chemotherapy (adjusted OR: 0.62; 95% confidence interval: 0.58-0.67; P < 0.001).

CONCLUSIONS

In a large, population-based analysis of brain metastasis patients, we found significant differences in treatment access and mild survival differences along socioeconomic strata. More specifically, patients in lower SES tiers suffered worse outcomes and received radiation and chemotherapy less frequently than patients in higher tiers, even after accounting for other tumor- and demographic-related information.

An Updated Comparison Between World Health Organization Grade II Gemistocytic and Diffuse Astrocytoma Survival and Treatment Patterns

BACKGROUND

In 2016, the World Health Organization revised its guidelines to retain only gemistocytic astrocytoma (GemA) as a distinct variant of diffuse astrocytoma (DA). In the past, grade II GemAs were linked with a worse prognosis than DA. However, it is unclear how consistently the tumor subtype has been diagnosed over time. We used more recent data to compare outcomes between grade II GemA and DA.

METHODS

Patients with grade II DA and GemA were extracted from the Surveillance, Epidemiology, and End Results database between 1973 and 2016. Kaplan-Meier curves estimated survival differences across different eras, with a focus on patients diagnosed between 2000 and 2016, and propensity score matching was used to balance baseline characteristics between DA and GemA cohorts.

RESULTS

Of 2467 patients with grade II astrocytoma diagnosed between 2000 and 2016, 132 (5.35%) had GemA, and 2335 (94.65%) had DA. At baseline, marked demographic and treatment differences were noted between tumor subtypes, including age at diagnosis and female sex. GemA patients did not have worse survival compared with DA patients at baseline (P = 0.349) or after propensity score matching (P = 0.497). Multivariate Cox models found that surgical extent of resection was associated with a survival benefit for DA patients, and both DA and GemA patients >65 years old had dramatically inferior survival.

CONCLUSIONS

Our data suggest that the impact of GemA versus DA histopathology depends more on the decade of queried data rather than patient-specific demographics. Using more recent longitudinal data, we found that grade II GemA and DA tumors did not have significant differences in survival. These data may prove useful for clinicians counseling patients with grade II GemA.

Differences in treatment patterns and overall survival between grade II and anaplastic pleomorphic xanthoastrocytomas

INTRODUCTION

Pleomorphic xanthoastrocytomas (PXAs) are classified as a grade II neoplasm, typically occur in children, and have favorable prognoses. However, their anaplastic counterparts remain poorly understood and vaguely characterized. In the present study, a large cohort of grade II PXA patients were compared with primary anaplastic PXA (APXA) patients to characterize patterns in treatment and survival.

METHODS

Data were collected from the National Cancer Institute's SEER database. Univariate and multivariate Cox regressions were used to evaluate the prognostic impact of demographic, tumor, and treatment-related covariates. Propensity score matching was used to balance baseline characteristics. Kaplan-Meier curves were used to estimate survival.

RESULTS

A total of 346 grade II PXA and 62 APXA patients were identified in the SEER database between 2000 and 2016. Kaplan-Meier analysis revealed substantially inferior survival for APXA patients compared to grade II PXA patients (median survival: 51 months vs. not reached) (p < 0.0001). After controlling across available covariates, increased age at diagnosis was identified as a negative predictor of survival for both grade II and APXA patients. In multivariate and propensity-matched analyses, extent of resection was not associated with improved outcomes in either cohort.

CONCLUSIONS

Using a large national database, we identified the largest published cohort of APXA patients to date and compared them with their grade II counterparts to identify patterns in treatment and survival. Upon multivariate analysis, we found increased age at diagnosis was inversely associated with survival in both grade II and APXA patients. Receipt of chemoradiotherapy or complete surgical resection was not associated with improved outcomes in the APXA cohort.

Patient Experience and Satisfaction with Telemedicine During Coronavirus Disease 2019: A Multi-Institution Experience

Introduction: The coronavirus disease 2019 (COVID-19) heralded an unprecedented increase in telemedicine utilization. Our objective was to assess patient satisfaction with telemedicine during the COVID-19 era. Methods: Telemedicine visit data were gathered from Stanford Health Care (Stanford) and the Hospital for Special Surgery (HSS). Patient satisfaction data from HSS were captured from a Press-Ganey questionnaire between April 19, 2020, and December 12, 2020, whereas Stanford data were taken from a novel survey instrument that was distributed to all patients between June 22, 2020, and November 1, 2020. Participants: There were 60,550 telemedicine visits at Stanford, each linked with a postvisit survey. At HSS, there were 66,349 total telemedicine visits with 7,348 randomly linked with a postvisit survey. Main Outcomes and Measures: Two measures of patient satisfaction were used for this study: (1) a patient's "overall visit score" and (2) whether the patient indicated the highest possible "likelihood to recommend" (LTR) score (LTR top box score). Results: The LTR top box percentage at Stanford increased from 69.6% to 74.0% (p = 0.0002), and HSS showed no significant change (p = 0.7067). In the multivariable model, the use of a cell phone (adjusted odds ratio [aOR]: 1.18; 95% confidence interval [CI]: 1.12-1.23) and tablet (aOR: 1.15; 95% CI: 1.07-1.23) was associated with higher overall scores, whereas visits with interrupted connections (aOR: 0.49; 95% CI: 0.42-0.57) or help required to connect (aOR: 0.49; 95% CI: 0.42-0.56) predicted lower patient satisfaction. Conclusions: We present the largest published description of patient satisfaction with telemedicine, and we identify important telemedicine-specific factors that predict increased overall visit score. These include the use of cell phones or tablets, phone reminders, and connecting before the visit was scheduled to begin. Visits with poor connectivity, extended wait times, or difficulty being seen, examined, or understood by the provider were linked with reduced odds of high scores. Our results suggest that attention to connectivity and audio/visual definition will help optimize patient satisfaction with future telemedicine encounters.

CMET-27. EVALUATION OF DYNAMIN 2 (DNM2) AS A THERAPEUTIC TARGET IN LEPTOMENINGEAL METASTATIC DISEASE

Leptomeningeal metastasis (LM), a spread of cancer to the meninges and cerebrospinal fluid, carries extremely poor prognosis due to fast progression and no effective treatment. Given that LM mostly develops in specific types of cancer, notably melanoma, breast and lung cancer, it is likely that predisposition for LMD is shaped by specific genetic footprints and/or anti-cancer therapies. In this regard, it is interesting that among all breast cancers the triple-negative (i.e. estrogen and progesterone receptor-negative and HER2 overexpression-negative) breast cancer (TNBC) type develops a disproportionally high percentage of LM, accounting for the majority (~40%) of all breast cancer LM cases. TNBCs trace their origin to the genomic instability stemming from defects in DNA repair (notably homology-directed repair, HDR). We have recently shown that the efficiency of HDR depends on dynamin 2 (DNM2) best known for its role in endocytic protein trafficking. Higher DNM2 was associated with more efficient HDR and the resistance to DNA-crosslinking chemotherapy. Importantly, elevated DNM2 was associated with lower relapse-free survival and shorter times to relapse after chemotherapy only in TNBCs and not in other types of breast cancer. As DNM2 also fuels migration and invasion, the cells with high DNM2 are not only the most resistant to chemotherapy but also are the most mobile, and thus may represent the core population of LM. Here we test the inhibition of the DNM2-dependent endocytic trafficking as a potential therapeutic strategy to halt LM in TNBC. As a model of LM we use the human-in-mouse model of brain metastasis based on internal carotid injection of MDA-MB-231-BR3 cells, which we have shown to faithfully recapitulate human LM. We show that DNM2 knockdown delays metastatic spread to the brain and potentiates the effect of DNA-crosslinking chemotherapeutic cyclophosphamide, providing justification for further testing of DNM2 inhibitors for targeted therapy of LM.