Real-world Evidence to Estimate Prostate Cancer Costs for First-line Treatment or Active Surveillance

Real-world utilization, patient characteristics, and treatment patterns among men with localized prostate cancer tested with the 17-gene genomic prostate score® (GPSTM) assay

OBJECTIVES

Descriptive study focusing on real-world utilization and characteristics of men with prostate cancer tested with the 17-gene Genomic Prostate Score® (GPS™) assay by linking administrative claims and electronic health record (EHR) data with GPS results.

METHODS

This retrospective, observational cohort study (January 1, 2013 to December 31, 2020) included men aged 40-80 years with localized prostate cancer claims, continuous enrollment in Optum's Integrated Claims data set, ≥1 day of EHR clinical activity, and a GPS result. Men were classified as undergoing definitive therapy (DT) (prostatectomy, radiation, or focal therapy) or active surveillance (AS). AS and DT distribution were analyzed across GPS results, National Comprehensive Cancer Network® (NCCN®) risk, and race. Costs were assessed 6 months after the first GPS result (index); clinical outcomes and AS persistence were assessed during the variable follow-up. All variables were analyzed descriptively.

RESULTS

Of 834 men, 650 (77.9%) underwent AS and 184 (22.1%) DT. Most men had Quan-Charlson comorbidity scores of 1-2 and a tumor stage of T1c (index). The most common Gleason patterns were 3 + 3 (79.6%) (AS cohort) and 3 + 4 (55.9%) (DT cohort). The mean (standard deviation) GPS results at index were 23.2 (11.3) (AS) and 30.9 (12.9) (DT). AS decreased with increasing GPS result and NCCN risk. Differences between races were minimal. Total costs were substantially higher in the DT cohort.

CONCLUSIONS

Most men with GPS-tested localized prostate cancer underwent AS, indicating the GPS result can inform clinical management. Decreasing AS with increasing GPS result and NCCN risk suggests the GPS complements NCCN risk stratification.

Active Surveillance for Taiwanese Men with Localized Prostate Cancer: Intermediate-Term Outcomes and Predictive Factors

PURPOSE

Active surveillance (AS) is one of the management options for patients with low-risk and select intermediate-risk prostate cancer (PC). However, factors predicting disease reclassification and conversion to active treatment from a large population of pure Asian cohorts regarding AS are less evaluated. This study investigated the intermediate-term outcomes of patients with localized PC undergoing AS.

MATERIALS AND METHODS

This cohort study enrolled consecutive men with localized non-high-risk PC diagnosed in Taiwan between June 2012 and Jan 2023. The study endpoints were disease reclassification (either pathological or radiographic progression) and conversion to active treatment. The factors predicting endpoints were evaluated using the Cox proportional hazards model.

RESULTS

A total of 405 patients (median age: 67.2 years) were consecutively enrolled and followed up with a median of 64.6 months. Based on the National Comprehensive Cancer Network (NCCN) risk grouping, 70 (17.3%), 164 (40.5%), 140 (34.6%), and 31 (7.7%) patients were classified as very low-risk, low-risk, favorable-intermediate risk, and unfavorable intermediate-risk PC, respectively. The 5-year reclassification rates were 24.8%, 27.0%, 18.6%, and 25.3%, respectively. The 5-year conversion rates were 20.4%, 28.8%, 43.6%, and 37.8%, respectively. A prostate-specific antigen density (PSAD) of ≥0.15 ng/mL² predicted reclassification (hazard ratio [HR] 1.84, 95% confidence interval [CI] 1.17-2.88) and conversion (HR 1.56, 95% CI 1.05-2.31). A maximal percentage of cancer in positive cores (MPCPC) of ≥15% predicted conversion (15% to <50%: HR 1.41, 95% CI 0.91-2.18; ≥50%: HR 1.97, 95% CI 1.1453-3.40) compared with that of <15%. A Gleason grade group (GGG) of 3 tumor also predicted conversion (HR 2.69, 95% CI 1.06-6.79; GGG 3 vs 1). One patient developed metastasis, but none died of PC during the study period (2,141 person-years).

CONCLUSIONS

AS is a viable option for Taiwanese men with non-high-risk PC, in terms of reclassification and conversion. High PSAD predicted reclassification, whereas high PSAD, MPCPC, and GGG predicted conversion.

Applying Natural Language Processing to Textual Data From Clinical Data Warehouses: Systematic Review

BACKGROUND

In recent years, health data collected during the clinical care process have been often repurposed for secondary use through clinical data warehouses (CDWs), which interconnect disparate data from different sources. A large amount of information of high clinical value is stored in unstructured text format. Natural language processing (NLP), which implements algorithms that can operate on massive unstructured textual data, has the potential to structure the data and make clinical information more accessible.

OBJECTIVE

The aim of this review was to provide an overview of studies applying NLP to textual data from CDWs. It focuses on identifying the (1) NLP tasks applied to data from CDWs and (2) NLP methods used to tackle these tasks.

METHODS

This review was performed according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. We searched for relevant articles in 3 bibliographic databases: PubMed, Google Scholar, and ACL Anthology. We reviewed the titles and abstracts and included articles according to the following inclusion criteria: (1) focus on NLP applied to textual data from CDWs, (2) articles published between 1995 and 2021, and (3) written in English.

RESULTS

We identified 1353 articles, of which 194 (14.34%) met the inclusion criteria. Among all identified NLP tasks in the included papers, information extraction from clinical text (112/194, 57.7%) and the identification of patients (51/194, 26.3%) were the most frequent tasks. To address the various tasks, symbolic methods were the most common NLP methods (124/232, 53.4%), showing that some tasks can be partially achieved with classical NLP techniques, such as regular expressions or pattern matching that exploit specialized lexica, such as drug lists and terminologies. Machine learning (70/232, 30.2%) and deep learning (38/232, 16.4%) have been increasingly used in recent years, including the most recent approaches based on transformers. NLP methods were mostly applied to English language data (153/194, 78.9%).

CONCLUSIONS

CDWs are central to the secondary use of clinical texts for research purposes. Although the use of NLP on data from CDWs is growing, there remain challenges in this field, especially with regard to languages other than English. Clinical NLP is an effective strategy for accessing, extracting, and transforming data from CDWs. Information retrieved with NLP can assist in clinical research and have an impact on clinical practice.

Metabolic signature biomarkers for predicting the recurrence of urological cancers

A significant number of patients diagnosed with localized urological cancers experience relapse and disease progression after surgery. Hence, molecular markers for patient risk stratification are needed to improve the current management guidelines. This article critically reviews for the first time, to our knowledge, the promise of metabolomics-based approaches to identify metabolic signatures as candidate prognostic biomarkers to predict recurrences at the time of surgery in prostate cancer (PCa), bladder cancer (BCa), and renal cell carcinoma (RCC). Dysregulations in the levels of several tumoral, circulating, and excreted metabolites have been reported in PCa patients experiencing recurrence within 1.5 to 8 years of follow-up. The combination of these metabolic biomarkers with clinical parameters (e.g., pathological T stage, Gleason score) has shown great potential to improve the predictive ability of PCa recurrence. In contrast, predictive biomarkers of recurrence in BCa and RCC have been poorly explored. Overall, this review highlights the great potential of metabolomics in discovering prognostic biomarkers for a more accurate patient risk stratification in urological cancers.

Health Care Costs Attributable to Prostate Cancer in British Columbia, Canada: A Population-Based Cohort Study

We aimed to estimate the total health care costs attributable to prostate cancer (PCa) during care phases by age, cancer stage, tumor grade, and primary treatment in the first year in British Columbia (BC), Canada. Using linked administrative health data, we followed a cohort of men aged ≥ 50 years at diagnosis with PCa between 2010 and 2017 (Cohort 1) from the diagnosis date until the date of death, the last date of observation, or 31 December 2019. Patients who died from PCa after 1 January 2010, were selected for Cohort 2. PCa attributable costs were estimated by comparing costs in patients to matched controls. Cohort 1 (n = 22,672) had a mean age of 69.9 years (SD = 8.9) and a median follow-up time of 5.2 years. Cohort 2 included 6942 patients. Mean PCa attributable costs were the highest during the first year after diagnosis ($14,307.9 [95% CI: $13,970.0, $14,645.8]) and the year before death ($9959.7 [$8738.8, $11,181.0]). Primary treatment with radiation therapy had significantly higher costs each year after diagnosis than a radical prostatectomy or other surgeries in advanced-stage PCa. Androgen deprivation therapy (and/or chemotherapy) had the highest cost for high-grade and early-stage cancer during the three years after diagnosis. No treatment group had the lowest cost. Updated cost estimates could inform economic evaluations and decision-making.

Comparison of the medical costs between active surveillance and other treatments for early prostate cancer in Japan using data from the PRIAS-JAPAN study

OBJECTIVES

To compare the medical costs of active surveillance with those of robot-assisted laparoscopic prostatectomy, brachytherapy, intensity-modulated radiation therapy, and hormone therapy for low-risk prostate cancer.

METHODS

The costs of protocol biopsies performed in the first year of surveillance (between January 2010 and June 2020) and those of brachytherapy and radiation therapy performed between May 2019 and June 2020 at the Kagawa University Hospital were analyzed. Hormone therapy costs were assumed to be the costs of luteinizing hormone-releasing hormone analogs for over 5 years. Active surveillance-eligible patients were defined based on the following: age <74 years, ≤T2, Gleason score ≤6, prostate-specific antigen level ≤10 ng/ml, and 1-2 positive cores. We estimated the total number of active surveillance-eligible patients in Japan based on the Japan Study Group of Prostate Cancer (J-CAP) study and the 2017 cancer statistical data. We then calculated the 5-year treatment costs of active surveillance-eligible patients using the J-CAP and PRIAS-JAPAN study data.

RESULTS

In 2017, number of active surveillance-eligible patients in Japan was estimated to be 2808. The 5-year total costs of surveillance, prostatectomy, brachytherapy, radiation therapy, and hormone therapy were 1.65, 14.0, 4.61, 4.04, and 5.87 million United States dollar (USD), respectively. If 50% and 100% of the patients in each treatment group had opted for active surveillance as the initial treatment, the total treatment cost would have been reduced by USD 6.89 million (JPY 889 million) and USD 13.8 million (JPY 1.78 billion), respectively.

CONCLUSION

Expanding active surveillance to eligible patients with prostate cancer helps save medical costs.

Expanding the Secondary Use of Prostate Cancer Real World Data: Automated Classifiers for Clinical and Pathological Stage

Background

Explicit documentation of stage is an endorsed quality metric by the National Quality Forum. Clinical and pathological cancer staging is inconsistently recorded within clinical narratives but can be derived from text in the Electronic Health Record (EHR). To address this need, we developed a Natural Language Processing (NLP) solution for extraction of clinical and pathological TNM stages from the clinical notes in prostate cancer patients.

Methods

Data for patients diagnosed with prostate cancer between 2010 and 2018 were collected from a tertiary care academic healthcare system's EHR records in the United States. This system is linked to the California Cancer Registry, and contains data on diagnosis, histology, cancer stage, treatment and outcomes. A randomly selected sample of patients were manually annotated for stage to establish the ground truth for training and validating the NLP methods. For each patient, a vector representation of clinical text (written in English) was used to train a machine learning model alongside a rule-based model and compared with the ground truth.

Results

A total of 5,461 prostate cancer patients were identified in the clinical data warehouse and over 30% were missing stage information. Thirty-three to thirty-six percent of patients were missing a clinical stage and the models accurately imputed the stage in 21–32% of cases. Twenty-one percent had a missing pathological stage and using NLP 71% of missing T stages and 56% of missing N stages were imputed. For both clinical and pathological T and N stages, the rule-based NLP approach out-performed the ML approach with a minimum F1 score of 0.71 and 0.40, respectively. For clinical M stage the ML approach out-performed the rule-based model with a minimum F1 score of 0.79 and 0.88, respectively.

Conclusions

We developed an NLP pipeline to successfully extract clinical and pathological staging information from clinical narratives. Our results can serve as a proof of concept for using NLP to augment clinical and pathological stage reporting in cancer registries and EHRs to enhance the secondary use of these data.