JCO Clinical Cancer Informatics最新文献

Purpose: Information on concerns that young adults (YAs) with cancer face when receiving care outside of specialized treatment centers is needed to increase equitable care to YAs at greater risk of marginalization by the health care system. The current study compared distress and unmet needs at the time of clinic visit between YAs receiving care from three different cancer clinics: (1) a National Cancer Institute-designated center, (2) a community-based clinic, and (3) a county hospital outpatient clinic.

Methods: The Adolescent and Young Adult Psycho-Oncology Screening Tool (AYA-POST) was administered to measure distress and cancer-related concerns of YAs in active treatment. A one-way analysis of variance (ANOVA) compared distress scores by treatment site. A Fisher's exact test compared the number of participants endorsing each item on the Needs Assessment Checklist from each site. A simple linear regression determined the association between distress and number of items endorsed on the Needs Assessment Checklist.

Results: Ninety-seven participants completed the AYA-POST, endorsing, on average, 11 concerns. Fisher's exact test showed significant differences between sites in the proportion of participants endorsing eight items: boredom (P < .001), eating/appetite (P < .001), nausea/vomiting (P < .001), financial concern (P = .002), hopelessness/helplessness (P = .03), confidentiality (P = .04), sibling concern (P = .04), and insurance (P = .05). The simple linear regression model was significant (F(1, 94) = 39.772, P < .001, R² = 0.297), indicating the number of unmet needs accounted for almost 30% of the variance in distress. The one-way ANOVA was not significant (F(2, 93) = 1.34, P = .267).

Conclusion: Social determinants of health can influence the number and type of unmet needs experienced, affecting distress and other outcomes and underscoring the importance of timely, effective, age-appropriate screening and intervention for distress and unmet needs in YAs with cancer.

Purpose: Real-world lung cancer data in administrative claims databases often lack staging information and specific diagnostic codes for lung cancer histology subtypes. This study updates and validates Turner's 2017 treatment-based algorithm using more recent claims and electronic health record (EHR) data.

Methods: This study used Optum's deidentified Market Clarity Data of linked medical and pharmacy claims with EHR data. Eligible patients had an incident lung cancer diagnosis (January 2014-December 2020) and ≥one valid histology code for lung cancer 30 days before to 60 days after diagnosis. Histology and stage information from the EHR were used to evaluate the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). We evaluated the Turner algorithm using cohort 1 patients diagnosed between June 2014 and October 2015 (step 1) and between November 2015 and December 2020 after approval of immunotherapies (step 2). Next, we evaluated cohort 2 patients diagnosed between November 2015 and December 2020 using an updated algorithm incorporating the latest US treatment guidelines (step 3), and compared the results for cohort 2 (Turner algorithm, step 2 patients). Furthermore, an algorithm to determine early NSCLC (eNSCLC; stage I-III) versus metastatic or advanced/metastatic non-small cell lung cancer (stage IV) was evaluated among patients with available histology and stage information.

Results: A total of 5,012 patients were included (cohort 1, step 1: n = 406; cohort 1, step 2: n = 2,573; cohort 2, step 3: n = 2,744). The updated algorithm showed improved performance relative to the previous Turner algorithm for sensitivity (0.920-0.932), specificity (0.865-0.923), PPV (0.976-0.988), and NPV (0.640-0.673). The eNSCLC algorithm showed high specificity (0.874) and relatively low sensitivity (0.539).

Conclusion: An updated treatment-based algorithm identifying patients with incident NSCLC was validated using EHR data and distinguished lung cancer subtypes in claims databases when EHR data were not available.

Purpose: The gender, age, lens culinaris agglutinin-reactive fraction of alphafetoprotein, alphafetoprotein, des-gamma-carboxyprothrombin (GALAD) score is a biomarker-based statistical model for the serologic diagnosis of hepatocellular carcinoma (HCC) that has been developed and validated using the case-control approach with a view to early detection. Performance has, however, been suboptimal in the first prospective studies which better reflect the real-world situation. In this article, we report the application of machine learning to a large, prospectively accrued, HCC surveillance data set.

Patients and methods: Models were built on a cohort of 3,473 patients with chronic liver disease within a rigorous surveillance program between 1998 and 2014, during which 459 patients with HCC were detected. Two random forest (RF) models were trained. The first RF model uses the same variables as the original GALAD model (GALAD-RF); the second is based on routinely available clinical and laboratory features (RF-practical). For comparison, we evaluated a logistic regression GALAD model trained on this longitudinal prospective data set (termed GALAD-Ogaki).

Results: Models were evaluated using a repetitive cross-validation approach with the metrics averaged over 100 independent runs. As judged by area under the receiver operator curve (AUROC) and F1 score, the GALAD RF model significantly outperformed the original GALAD model. The RF-practical model also outperformed the original GALAD model in terms of both AUROC and F1 score, and both models outperformed the individual biomarkers. An online web application that implemented the GALAD-RF and RF-practical models is presented.

Conclusion: RF-based models improve on the diagnostic performance of the original GALAD model in the setting of a standard HCC surveillance program. Further prospective validation studies are warranted using these models and could be expanded to offer prediction of risk of HCC development over defined periods of time.

Standardizing image-data preparation practices to improve accuracy/consistency of AI diagnostic tools.

Purpose: Ability to predict symptom severity and progression across treatment trajectories would allow clinicians to provide timely intervention and treatment planning. However, such predictions are difficult because of sparse and inconsistent assessment, and simplistic measures such as the last observed symptom severity are often used. The purpose of this study is to develop a model for predicting future cancer symptom experiences on the basis of past symptom experiences.

Patients and methods: We performed a retrospective, longitudinal analysis using records of patients with cancer (n = 208) hospitalized between 2008 and 2014. A long short-term memory (LSTM)-based recurrent neural network, a linear regression, and random forest models were trained on previous symptoms experienced and used to predict future symptom trajectories.

Results: We found that at least one of three tested models (LSTM, linear regression, and random forest) outperform predictions based solely on the previous clinical observation. LSTM models significantly outperformed linear regression and random forest models in predicting nausea (P < .1) and psychosocial status (P < .01). Linear regression outperformed all models when predicting oral health (P < .01), while random forest outperformed all models when predicting mobility (P < .01) and nutrition (P < .01).

Conclusion: We can successfully predict patients' symptom trajectories with a prediction model, built with sparse assessment data, using routinely collected nursing documentation. The results of this project can be applied to better individualize symptom management to support cancer patients' quality of life.

New publication provides insights into the impact of disability on outcomes in older adults with multiple myeloma.

Purpose: In a previous exploratory study, modeled early longitudinal prostate-specific antigen (PSA) kinetics observed within the 100-first treatment days with androgen deprivation therapy with or without docetaxel was associated with progression-free survival (PFS) and overall survival (OS) in patients with prostate cancer with rising PSA levels after primary local therapy. This prognostic value had to be confirmed in different settings. The objectives were to assess PSA kinetics modeling in patients with metastatic castration-resistant prostate cancer (mCRPC) treated with chemotherapy in FIRSTANA trial and to investigate modeled PSA kinetic parameters prognostic/predictive value.

Materials and methods: FIRSTANA phase III trial (ClinicalTrials.gov identifier: NCT01308567) assessed whether cabazitaxel is superior to docetaxel in terms of PFS/OS in patients with chemotherapy-naïve mCRPC. PSA longitudinal kinetics was assessed using the previous kinetic-pharmacodynamics model. Patient modeled ELIMination rate constant K (PSA.KELIM) was used to categorize favorable/unfavorable PSA declines (standardized PSA.KELIM < or ≥ 1.0 days^-1) and further correlated with PFS/OS.

Results: In total, 1,050 of 1,168 enrolled patients were assessable for PSA.KELIM estimation. The median PSA.KELIM was 0.02 days^-1. In univariate analyses, PSA.KELIM exhibited a significant prognostic value regarding survival: unfavorable versus favorable PSA.KELIM; median PFS, 3.6 months (95% CI, 3.0 to 4.2) versus 4.7 months (95% CI, 3.9 to 5.2), P = .002; median OS, 17.4 months (95% CI, 14.8 to 19.3) versus 28.4 months (95% CI, 26.7 to 31.6), P < .001. In multivariate analyses, PSA.KELIM was significant for PFS (hazard ratio [HR], 0.79 [95% CI, 0.67 to 0.93], P = .005) and OS (HR, 0.51 [95% CI, 0.44 to 0.60], P < .001), together with baseline radiological tumor progression and PSA doubling time. PSA.KELIM predictive value was not significant across treatment arms.

Conclusion: This external validation study confirmed previous results about modeled PSA longitudinal kinetics prognostic value regarding PFS/OS in patients with mCRPC treated with taxanes. PSA.KELIM could be used to identify a subpopulation with poor prognosis, who may benefit from treatment intensification.

Purpose: Although immune checkpoint inhibitors (ICIs) have improved outcomes in certain patients with cancer, they can also cause life-threatening immunotoxicities. Predicting immunotoxicity risks alongside response could provide a personalized risk-benefit profile, inform therapeutic decision making, and improve clinical trial cohort selection. We aimed to build a machine learning (ML) framework using routine electronic health record (EHR) data to predict hepatitis, colitis, pneumonitis, and 1-year overall survival.

Methods: Real-world EHR data of more than 2,200 patients treated with ICI through December 31, 2018, were used to develop predictive models. Using a prediction time point of ICI initiation, a 1-year prediction time window was applied to create binary labels for the four outcomes for each patient. Feature engineering involved aggregating laboratory measurements over appropriate time windows (60-365 days). Patients were randomly partitioned into training (80%) and test (20%) sets. Random forest classifiers were developed using a rigorous model development framework.

Results: The patient cohort had a median age of 63 years and was 61.8% male. Patients predominantly had melanoma (37.8%), lung cancer (27.3%), or genitourinary cancer (16.4%). They were treated with PD-1 (60.4%), PD-L1 (9.0%), and CTLA-4 (19.7%) ICIs. Our models demonstrate reasonably strong performance, with AUCs of 0.739, 0.729, 0.755, and 0.752 for the pneumonitis, hepatitis, colitis, and 1-year overall survival models, respectively. Each model relies on an outcome-specific feature set, though some features are shared among models.

Conclusion: To our knowledge, this is the first ML solution that assesses individual ICI risk-benefit profiles based predominantly on routine structured EHR data. As such, use of our ML solution will not require additional data collection or documentation in the clinic.

Purpose: Surgical pathology reports are critical for cancer diagnosis and management. To accurately extract information about tumor characteristics from pathology reports in near real time, we explore the impact of using domain-specific transformer models that understand cancer pathology reports.

Methods: We built a pathology transformer model, Path-BigBird, by using 2.7 million pathology reports from six SEER cancer registries. We then compare different variations of Path-BigBird with two less computationally intensive methods: Hierarchical Self-Attention Network (HiSAN) classification model and an off-the-shelf clinical transformer model (Clinical BigBird). We use five pathology information extraction tasks for evaluation: site, subsite, laterality, histology, and behavior. Model performance is evaluated by using macro and micro F₁ scores.

Results: We found that Path-BigBird and Clinical BigBird outperformed the HiSAN in all tasks. Clinical BigBird performed better on the site and laterality tasks. Versions of the Path-BigBird model performed best on the two most difficult tasks: subsite (micro F₁ score of 72.53, macro F₁ score of 35.76) and histology (micro F₁ score of 80.96, macro F₁ score of 37.94). The largest performance gains over the HiSAN model were for histology, for which a Path-BigBird model increased the micro F₁ score by 1.44 points and the macro F₁ score by 3.55 points. Overall, the results suggest that a Path-BigBird model with a vocabulary derived from well-curated and deidentified data is the best-performing model.

Conclusion: The Path-BigBird pathology transformer model improves automated information extraction from pathology reports. Although Path-BigBird outperforms Clinical BigBird and HiSAN, these less computationally expensive models still have utility when resources are constrained.

Purpose: Multiple myeloma (MM) is a prevalent hematologic malignancy in older adults, who often experience physical disability, increased health care usage, and reduced treatment tolerance. Home health (HH) services are frequently used by this group, but the relationship between disability, HH use, and MM treatment receipt is unclear. This study examines the connections between disability, treatment receipt, and survival outcomes in older adults with newly diagnosed MM using a nationwide data set.

Methods: The SEER-Medicare data set was used to identify adults aged 66 years and older diagnosed with MM from 2010 to 2017, who used HH services the year before diagnosis. Disability was assessed with the Outcome and Assessment Information Set, using a composite score derived from items related to ability to complete activities of daily living. Mortality, therapy receipt, and health care utilization patterns were evaluated.

Results: Of 37,280 older adults with MM, 6,850 (18.2%) used HH services before diagnosis. Moderate disability at HH assessment resulted in similar MM-directed therapy receipt as mild disability, with comparable health care usage after diagnosis to severe disability. HH users had a higher comorbidity burden and higher mortality (adjusted risk ratio for 3-year mortality: 1.59 [95% CI, 1.55 to 1.64]). Severe functional disability before diagnosis was strongly related to postdiagnosis mortality.

Conclusion: Among older adults with MM receiving HH services, disability is a predictor of early mortality. Moderately disabled individuals undergo similar therapy intensity as the mildly disabled but experience increased acute care utilization. Previous HH use could identify patients with MM requiring intensive support during therapy initiation.