Developing Artificial Intelligence Models for Extracting Oncologic Outcomes from Japanese Electronic Health Records.

INTRODUCTION: A framework that extracts oncological outcomes from large-scale databases using artificial intelligence (AI) is not well established. Thus, we aimed to develop AI models to extract outcomes in patients with lung cancer using unstructured text data from electronic health records of multiple hospitals. METHODS: We constructed AI models (Bidirectional Encoder Representations from Transformers [BERT], Naïve Bayes, and Longformer) for tumor evaluation using the University of Miyazaki Hospital (UMH) database. This data included both structured and unstructured data from progress notes, radiology reports, and discharge summaries. The BERT model was applied to the Life Data Initiative (LDI) data set of six hospitals. Study outcomes included the performance of AI models and time to progression of disease (TTP) for each line of treatment based on the treatment response extracted by AI models. RESULTS: For the UMH data set, the BERT model exhibited higher precision accuracy compared to the Naïve Bayes or the Longformer models, respectively (precision [0.42 vs. 0.47 or 0.22], recall [0.63 vs. 0.46 or 0.33] and F1 scores [0.50 vs. 0.46 or 0.27]). When this BERT model was applied to LDI data, prediction accuracy remained quite similar. The Kaplan-Meier plots of TTP (months) showed similar trends for the first (median 14.9 [95% confidence interval 11.5, 21.1] and 16.8 [12.6, 21.8]), the second (7.8 [6.7, 10.7] and 7.8 [6.7, 10.7]), and the later lines of treatment for the predicted data by the BERT model and the manually curated data. CONCLUSION: We developed AI models to extract treatment responses in patients with lung cancer using a large EHR database; however, the model requires further improvement.

The use of artificial intelligence (AI) to derive health outcomes from large electronic health records is not well established. Thus, we built three different AI models: Bidirectional Encoder Representations from Transformers (BERT), Naïve Bayes, and Longformer to serve this purpose. Initially, we developed these models based on data from the University of Miyazaki Hospital (UMH) and later improved them using the Life Data Initiative (LDI) data set of six hospitals. The performance of the BERT model was better than the other two, and it showed similar results when it was applied to the LDI data set. The Kaplan­Meier plots of time to progression of disease for the predicted data by the BERT model showed similar trends to those for the manually curated data. In summary, we developed an AI model to extract health outcomes using a large electronic health database in this study; however, the performance of the AI model could be improved using more training data.

Treatment status and safety of crizotinib in 2028 Japanese patients with ALK-positive NSCLC in clinical settings.

OBJECTIVE: Post-marketing surveillance (PMS) was performed in Japan to obtain information on the safety and efficacy of crizotinib. METHODS: Target patients included almost all patients with anaplastic lymphoma kinase-positive non-small cell lung cancer who were administered crizotinib. The observation period was 52 weeks. In the present study, we focused on the treatment status and safety of crizotinib therapy and analyzed the real-world data obtained by this PMS (ClinicalTrials.gov: NCT01597258). RESULTS: The safety analysis set included 2028 Japanese patients, and more than half of the patients (56.4%) were nonsmokers. The incidence of adverse drug reactions (ADRs) was 91.6%, and common ADRs (incidence ≥15%) were nausea (32.2%), diarrhea (24.3%), photopsia (18.9%), vomiting (17.5%) and dysgeusia (16.8%). Many patients (623 patients) discontinued treatment of crizotinib because of adverse events within 12 weeks after therapy initiation, which tended to frequently occur in the following cases: (1) elderly, (2) body weight <40 kg, (3) body surface area <1.2 m2 (4) ECOG PS 2-4, (5) higher Brinkman index and (6) history of occupational/environmental exposure such as asbestos/pneumoconiosis. The proportions of patients remaining on crizotinib therapy were 68.2% for 3 months, 55.2% for 6 months and 36.1% for 12 months, with a median duration of 7.9 months. Multivariate analysis with a Cox proportional hazard model identified 10 statistically significant patient background factors influencing the duration of crizotinib therapy. CONCLUSIONS: No new safety concerns were observed in this PMS study. Our results provide useful information regarding the status of crizotinib therapy in the clinical setting.

Efficacy and safety of sunitinib in Japanese patients with progressive, advanced/metastatic, well-differentiated, unresectable pancreatic neuroendocrine tumors: final analyses from a Phase II study.

BACKGROUND: In an interim analysis of a Phase II trial in Japanese patients with pancreatic neuroendocrine tumors (panNETs), sunitinib demonstrated antitumor activity with an objective response rate (ORR) of 50% (95% confidence interval [CI], 21-79) and a median progression-free survival (PFS) of 16.8 months (95% CI, 9.3-26.2). Here, we report the final analyses of efficacy and safety, as well as additional analyses, from this Phase II study. METHODS: This was a multicenter, open-label, Phase II trial (NCT01121562) of sunitinib in Japanese patients with panNETs. Patients received oral sunitinib 37.5 mg/day on a continuous daily dosing schedule. Dose modifications were permitted. The primary endpoint was clinical benefit rate (CBR). Secondary endpoints included ORR, PFS, overall survival (OS), safety and pharmacokinetics. RESULTS: Of 12 patients enrolled and treated, all discontinued treatment-the majority (n = 8) owing to disease progression. Most patients were male (n = 8), <65 years of age (n = 11) and had a non-functional tumor (n = 10). The median (range) number of days on drug was 323.5 (22-727). The CBR (95% CI) was 75.0% (42.8-94.5). ORR (95% CI) was 50.0% (21.1-78.9). Median (95% CI) PFS was 16.8 (9.3-26.2) months; however, median (95% CI) OS was not reached (22.0-not estimable). Most common adverse events (AEs; all-causality) were diarrhea (n = 10; 83.3%), hand-foot syndrome (n = 8; 66.7%) and hypertension (n = 8; 66.7%). CONCLUSIONS: These results support the efficacy and safety of sunitinib in Japanese patients with panNETs. Appropriate AE management through dose reduction and interruption may prolong sunitinib treatment and maximize its efficacy.

Correction to: Phase II study of sunitinib in Japanese patients with unresectable or metastatic, well-differentiated pancreatic neuroendocrine tumor.

In the original publication of this article, the license subtype should be CC BY and not CC BY-NC.

Real-world use of sunitinib in Japanese patients with pancreatic neuroendocrine tumors: results from a post-marketing surveillance study.

BACKGROUND: Sunitinib is approved for the treatment of progressive, well-differentiated pancreatic neuroendocrine tumors (pNETs) in patients with unresectable, locally advanced or metastatic disease. Safety and efficacy data in Japanese patients are limited. We report outcomes from a post-marketing surveillance study of sunitinib treatment in Japanese patients. METHODS: Sunitinib 37.5 mg once daily was orally administered in Japanese patients aged ≥ 15 years with pNETs. The primary endpoints included adverse events (AEs) occurring during the observation period of 168 days and objective response rate (ORR). RESULTS: Sunitinib was administered in 62 patients with pNETs. The median duration of treatment was 165 days. At 168 days from the start of treatment, 31 patients were still receiving sunitinib treatment and treatment continuation rate was 50.0%. Of the 31 patients who discontinued treatment, 18 (58.1%) discontinued because of AEs and 16 (51.6%) patients discontinued due to insufficient clinical effect. Of the 18 patients who discontinued due to AEs, 10 did so within 42 days of treatment initiation. The most common all-grade AEs were platelet count decreased (33.9%), diarrhea (29.0%), neutrophil count decreased (27.4%), hypertension (24.2%), and palmar-plantar erythrodysesthesia syndrome (24.2%). In the 51 patients eligible for the efficacy analysis, ORR was 13.7% (95% confidence interval, 5.7-26.3) and clinical benefit rate was 70.6%. CONCLUSIONS: There were no new safety concerns in real-world use of sunitinib in Japanese patients with pNETs. The short treatment duration likely led to low tumor response. Appropriate AEs management through dose interruption/reduction is essential for sunitinib treatment success in this patient population.

Interstitial Lung Disease Onset and Its Risk Factors in Japanese Patients With ALK-Positive NSCLC After Treatment With Crizotinib.

INTRODUCTION: The study objective was to determine the incidence and characteristics of drug-induced interstitial lung disease (ILD) associated with an orally available small-molecule tyrosine kinase inhibitor, crizotinib, in a real-world clinical setting. METHODS: Post-marketing surveillance was performed in Japan to obtain information on the safety and efficacy of crizotinib. Target patients included all patients with anaplastic lymphoma kinase-positive NSCLC who received crizotinib during the enrollment period between May 2012 and December 2014. The observation period was 52 weeks. Expert analysis of the ILD incidence was performed by an ILD independent review committee composed of five medical specialists. RESULTS: The safety analysis set included 2028 patients, and more than half of the patients (56.4%) were nonsmokers. The incidence of ILD associated with crizotinib therapy was 5.77%; and 3.45% patients showed grade 3 or greater. Pulmonary edema-like shadows with or without diffuse alveolar damage pattern were observed in crizotinib-associated ILD (incidence: 0.39%), but a causal relationship with the prognosis could not be identified. ILD developed within 4 weeks from initiation of crizotinib administration in 41.9% and within 8 weeks in 69.2% of the patients. Age 55 years or older, Eastern Cooperative Oncology Group performance status 2-4, smoking history, previous or concomitant ILD, and comorbid pleural effusion were statistically determined as significant risk factors for crizotinib-induced ILD. CONCLUSIONS: Crizotinib therapy should be applied to the NSCLC patients with any of above risk factors under a cautious monitoring for ILD occurrence, and clinicians should pay attention to the risks of severe ILD.

Crizotinib versus Chemotherapy in Asian Patients with ALK-Positive Advanced Non-small Cell Lung Cancer.

PURPOSE: Crizotinib has demonstrated superior progression-free survival (PFS) and objective response rates (ORRs) versus chemotherapy in previously treated and untreated patients with anaplastic lymphoma kinase (ALK)-positive advanced non-small cell lung cancer (NSCLC). We report the safety and efficacy of crizotinib in Asian subpopulations of two global phase III trials. MATERIALS AND METHODS: This analysis evaluated previously treated and untreated patients in two randomized, openlabel phase III trials of crizotinib versus chemotherapy in ALK-positive advanced NSCLC in second-line (PROFILE 1007) and first-line settings (PROFILE 1014). Efficacy and safety were analyzed by race in the intention-to-treat and "as-treated" populations for efficacy and safety endpoints, respectively. RESULTS: In previously treated (n=157) and untreated (n=157) Asian patients, PFS was statistically significantly longer with crizotinib versus chemotherapy (hazard ratio for PFS, 0.526; 95% confidence interval, 0.363 to 0.762; p < 0.001 and hazard ratio, 0.442; 95% confidence interval, 0.302 to 0.648; p < 0.001, respectively). Similar antitumor activity was seen in the non-Asian and overall populations. ORRs were statistically significantly higher with crizotinib versus chemotherapy in both Asian and non-Asian previously treated and untreated patients (p < 0.05). The most common treatment-emergent adverse events (any grade)with crizotinib were vision disorder, diarrhea, and nausea, which were observed at a comparable incidence across Asian and non-Asian populations, irrespective of previous treatment status. Most adverse events were mild to moderate in severity. CONCLUSION: These data, currently the only analysis showing Asian and non-Asian populations in the same study, support the efficacy and safety of crizotinib in Asian patients with previously treated or untreated ALK-positive advanced NSCLC.

Safety, efficacy and prognostic analyses of sunitinib in the post-marketing surveillance study of Japanese patients with gastrointestinal stromal tumor.

OBJECTIVE: This study was conducted to expand the sunitinib safety database in Japanese imatinib-resistant/-intolerant gastrointestinal stromal tumor patients. Retrospective analyses investigated common adverse events as potential prognostic markers. METHODS: Four hundred and seventy patients who received sunitinib between June 2008 and November 2009 were analyzed for safety, progression-free survival and overall survival; 386 for objective response rate; 88% received sunitinib on Schedule 4/2 starting at 50 mg/day. RESULTS: No unexpected safety issues occurred. Grade ≥ 3 adverse events occurred in 70%, most commonly thrombocytopenia (33%), neutropenia (22%) and leukopenia (15%). Objective response rate was 20% (95% confidence interval 16-24). Median progression-free survival was 22.4 weeks (95% confidence interval, 21.7-24.0). The overall survival rate at 24 weeks was 91% (95% confidence interval, 88-94). Higher relative dose intensity (≥70 vs. <70%) during the first 6 weeks and better Eastern Cooperative Oncology Group performance status (0 vs. ≥1) were associated with longer progression-free survival (24.0 vs. 20.1 weeks; P = 0.011; and 24.1 vs. 16.9 weeks; P < 0.001) and higher 24-week overall survival rate (94 vs. 83%; P < 0.001; and 96 vs. 83%; P < 0.001). Increased progression-free survival and overall survival rates were associated with specific adverse events. Cox proportional hazard modeling adjusted for relative dose intensity and performance status established hand-foot syndrome (hazard ratio = 0.636; 95% confidence interval, 0.456-0.888) and leukopenia (hazard ratio = 0.683; 95% confidence interval, 0.492-0.948) occurring within 12 weeks were significantly correlated with increased progression-free survival. CONCLUSION: Sunitinib showed good efficacy and tolerable safety. Factors associated with greater efficacy were relative dose intensity, performance status and specific early adverse events.

Phase II study of sunitinib in Japanese patients with unresectable or metastatic, well-differentiated pancreatic neuroendocrine tumor.

BACKGROUND: Pancreatic neuroendocrine tumors (NETs) are rare but are frequently diagnosed at advanced stages and require systemic therapy. PATIENTS AND METHODS: This multicenter, open-label, phase II study evaluated sunitinib in Japanese patients with well-differentiated pancreatic NET. Patients received sunitinib 37.5 mg/day on a continuous daily dosing (CDD) schedule. The primary endpoint was clinical benefit rate (CBR; percentage of complete responses [CRs] plus partial responses [PRs] plus stable disease [SD] ≥ 24 weeks). Secondary endpoints included objective response rate (ORR), tumor shrinkage, progression-free survival (PFS) probability, safety, pharmacokinetics, and biomarkers. RESULTS: Twelve patients received treatment. The CBR was 75 % (95 % confidence interval [CI], 43-94) and included 6 patients with a PR and 3 with SD. The ORR was 50 % (95 % CI, 21-79). PFS probability was 91 % (95 % CI, 54-99) at 6 months and 71 % (95 % CI, 34-90) at 12 months. Commonly reported treatment-emergent (all-causality), any-grade adverse events included diarrhea (n=10), hand-foot syndrome and hypertension (both n=8), fatigue and headache (both n=7), and neutropenia (n=6). No deaths on study were reported; one death due to disease progression occurred >28 days after end of treatment. Sunitinib on a CDD schedule resulted in sustained drug concentrations without accumulation across cycles. Tumor responses in all 12 patients did not appear to correlate with decreases in chromogranin A levels. CONCLUSIONS: Sunitinib 37.5 mg/day on a CDD schedule demonstrated antitumor activity in Japanese patients with unresectable, well-differentiated pancreatic NET. Commonly reported adverse events were consistent with the known safety profile of sunitinib.

Phase I and pharmacokinetic study of dacomitinib (PF-00299804), an oral irreversible, small molecule inhibitor of human epidermal growth factor receptor-1, -2, and -4 tyrosine kinases, in Japanese patients with advanced solid tumors.

BACKGROUND: Dacomitinib (PF-00299804) is an oral, irreversible, small molecule inhibitor of human epidermal growth factor receptor-1, -2, and -4 tyrosine kinases. METHODS: This phase I, open-label, dose-escalation study (clinicaltrials.gov: NCT00783328) primarily evaluated the safety and tolerability of dacomitinib by dose-limiting toxicity (DLT), and determined the clinically recommended phase II dose (RP2D) in Japanese patients with advanced solid tumors. Dacomitinib was administered orally at three dose levels (15, 30, or 45 mg once daily [QD]). Patients initially received a single dose, and after 9 days of follow-up, continuously QD in 21-day cycles. Endpoints included pharmacokinetics (PK) and antitumor activity. RESULTS: Thirteen patients were assigned to the three dose levels (15 mg cohort: n = 3; 30 mg cohort: n = 3; 45 mg cohort: n = 7) according to a traditional '3 + 3' design. None of the treated patients experienced a DLT. Toxicities were manageable and similar in type to those observed in other studies. PK concentration parameters increased with dose over the range evaluated, with no evidence of accumulation over time. Of 13 evaluable patients, one with NSCLC (adenocarcinoma) had a partial response and nine patients had stable disease. CONCLUSIONS: Dacomitinib 45 mg QD was defined as the RP2D and demonstrated preliminary activity in Japanese patients with advanced solid tumors.

Feasibility study of two schedules of sunitinib in combination with pemetrexed in patients with advanced solid tumors.

BACKGROUND: Sunitinib is an oral multitargeted tyrosine kinase inhibitor of vascular endothelial growth factor and platelet-derived growth factor receptors, as well as of other receptor types. We have performed a feasibility study to investigate the safety of sunitinib in combination with pemetrexed for treatment of advanced refractory solid tumors. METHODS: Sunitinib was administered once daily on a continuous daily dosing (CDD) schedule (37.5 mg/day) or a 2-weeks-on, 1-week-off treatment schedule (50 mg/day, Schedule 2/1) in combination with pemetrexed at 500 mg/m(2) on day 1 of repeated 21-day cycles. RESULTS: Twelve patients were enrolled in the study: six on the CDD schedule and six on Schedule 2/1. None of the treated patients experienced a dose-limiting toxicity. Toxicities were manageable and similar in type to those observed in monotherapy studies of sunitinib and pemetrexed. Pharmacokinetic analysis did not reveal any substantial drug-drug interaction. One patient with squamous cell lung cancer showed a partial response and five patients had stable disease. CONCLUSIONS: Combination therapy with sunitinib administered on Schedule 2/1 (50 mg/day) or a CDD schedule (37.5 mg/day) together with standard-dose pemetrexed (500 mg/m(2)) was well tolerated in previously treated patients with advanced solid tumors.

Figitumumab combined with carboplatin and paclitaxel in treatment-naïve Japanese patients with advanced non-small cell lung cancer.

OBJECTIVES: The insulin-like growth factor (IGF) signaling pathway has been implicated in the pathogenesis of numerous tumor types, including non-small cell lung cancer (NSCLC). Figitumumab is a fully human IgG2 monoclonal antibody against IGF-1 receptor (IGF-1R). METHODS: This phase I, open-label, dose-escalation study (ClinicalTrials.gov: NCT00603538) assessed the safety and tolerability of figitumumab (6, 10 and 20 mg/kg) in combination with carboplatin (area under the curve: 6 mg·min/mL) and paclitaxel (200 mg/m(2)) in Japanese patients (N = 19) with chemotherapy-naïve, advanced NSCLC. Treatments were administered intravenously on day 1 of a 21-day cycle for four to six cycles. Pharmacokinetics, biomarkers, and antitumor activity were also evaluated. RESULTS: Figitumumab in combination with carboplatin and paclitaxel was well tolerated at doses up to 20 mg/kg; no dose-limiting toxicities were observed at this dose level. When given in combination, figitumumab plasma exposure increased in an approximately dose-proportional manner. The approximate 2-fold accumulation following repeated administration supported the 21-day regimen as appropriate for figitumumab administration. Serum total IGF-1 and IGF binding protein-3 concentrations increased following figitumumab dosing, but a clear dose-dependent relationship was not demonstrated. Seven of 18 evaluable patients experienced a partial response. CONCLUSIONS: Figitumumab 20 mg/kg in combination with carboplatin and paclitaxel was well tolerated in chemotherapy-naïve Japanese patients with NSCLC. Further analysis of biomarker data is necessary for the development of figitumumab therapy.

Phase I study of TLR9 agonist PF-3512676 in combination with carboplatin and paclitaxel in patients with advanced non-small-cell lung cancer.

This phase I, open-label study investigated the Toll-like receptor 9 agonist, PF-3512676, in combination with carboplatin and paclitaxel in Japanese patients with advanced, non-small-cell lung cancer (NSCLC). Patients (n = 12) with treatment-naive stage IIIB or IV NSCLC received single-agent PF-3512676 subcutaneously once during the first 7 days (monotherapy phase) in three escalating dose levels (0.1, 0.2, and 0.4 mg/kg) followed by a combination phase during which patients received 0.1 or 0.2 mg/kg PF-3512676 subcutaneously on days 8 and 15 of each 3-week cycle of carboplatin (area under the curve, 6 mg x min/mL) and paclitaxel (200 mg/m(2)). Safety and pharmacokinetics of PF-3512676 were assessed during monotherapy and combination therapy phases. PF-3512676 was tolerable as monotherapy or in combination with chemotherapy in patients with NSCLC. Most common treatment-related, non-hematologic adverse events (AEs) throughout the study were injection-site reactions (n = 12, 100%) and flu-like symptoms (n = 11, 91.7%) that were each grade 1 or 2 in all but one patient. All patients experienced neutropenia and leukopenia (>or=grade 3 in 11 [91.7%] and seven [58.3%] patients, respectively). One patient in dose level 2 had a dose-limiting toxicity: grade 3 rash and grade 3 increase in gamma-glutamyltransferase during combination therapy. Mean PF-3512676 half-life ranged from 4.8 to 21.6 h (longer with higher doses). Four (33%) patients had objective responses (one complete response, three partial responses), and seven (58%) patients achieved stable disease. PF-3512676 as monotherapy and in combination with chemotherapy had an acceptable safety profile in Japanese patients with treatment-naive NSCLC.