Administrative data algorithms to identify second breast cancer events following early-stage invasive breast cancer

Real-world perioperative treatment patterns and economic burden of recurrence in early-stage HER2-negative breast cancer: a SEER-Medicare study

AIM

This study aimed to describe treatment patterns and quantify the economic impact of recurrence in early-stage human epidermal growth factor receptor 2 (HER2)-negative breast cancer (BC).

MATERIALS & METHODS

Medicare beneficiaries with stages I-III HER2-negative BC and lumpectomy or partial/total mastectomy were identified from SEER-Medicare data (2010-2019). Perioperative therapies were reported in the neoadjuvant and adjuvant setting. Locoregional recurrence and distant metastasis were identified using a claims-based algorithm developed with clinical input and consisting of a diagnosis-based and treatment-based indicator. All-cause and BC-related healthcare resource utilization (HRU) per-patient-month and monthly healthcare costs were estimated from the recurrence date for patients with recurrence and from an imputed index date for patients without recurrence using frequency matching. HRU and costs were compared between groups stratified by hormone receptor-positive (HR+) or triple negative BC (TNBC) using multivariable regression models.

RESULTS

Of 28,655 patients, 8.5% experienced recurrence, 90.4% had HR+ disease, and 5.6% received neoadjuvant therapy. Relative to patients without recurrence, patients with recurrence had more advanced disease (stage II/III: 73.7% vs. 34.0%) and higher-grade tumors (Grade 3/4: 40.6% vs. 18.0%) at diagnosis. Recurrence in HR+/HER2-negative BC and TNBC was associated with higher rates of all-cause hospitalizations (incidence rate ratio [IRR]: 2.84 and 3.65), emergency department (ED) visits (IRR: 1.75 and 2.00), and outpatient visits (IRR: 1.46 and 1.55; all p < 0.001). Similarly, recurrence was associated with higher rates of BC-related HRU, particularly for ED visits in HR+/HER2-negative BC (IRR: 4.24; p < 0.001) and hospitalizations in TNBC (IRR: 11.71; p < 0.001). Patients with HR+/HER2-negative BC and TNBC recurrence incurred higher monthly all-cause (cost difference [CD]: $3988 and $4651) and BC-related healthcare costs (CD: $3743 and $5819).

CONCLUSIONS

Our findings highlight the considerable economic burden of recurrence in early-stage HER2-negative BC and underscore the unmet need for optimization of therapies that reduce recurrence in this population.

Incidence and temporal patterns of true recurrences and second primaries in women with breast cancer: A 10-year competing risk-adjusted analysis

INTRODUCTION

We report a population-based, competing risk-adjusted analysis of the risk and timing of true recurrences and second primaries in women with breast cancer (BC), that are still ill-defined.

METHODS

We performed a manual review of medical charts of 1988 BC patients from a cancer registry in northern Italy (2000-2013). The occurrence and timing of true recurrences (TRs, including local, regional and distant recurrences) and second BCs (SBCs, including ipsilateral and contralateral SBC) during 10 years of follow-up were evaluated. The prognostic factors for TRs and SBCs were identified using the Fine and Gray model.

RESULTS

The cumulative incidence was 13.7 % (95 % confidence interval (CI), 12.2-15.3 %) for TRs and 4.6 % (95 % CI, 3.7-5.7 %) for SBCs. The median time to detection varied between 3.4 (TRs) and 5.1 (SBCs) years. The risk of TRs had two peaks, one between the 2nd and the 3rd year of follow-up and another between the 7th and the 8th year. The subhazard of SBCs fluctuated for five years, had a drop between the 6th and the 7th year and a marked peak between the 8th and the 9th year. Prognostic factors for TRs (tumour stage and grade, lymph node status and residual disease) and SBCs (patient age and -inverse association- hormone therapy) were different. In the 9th-10th year of follow-up, the excess incidence of total BC episodes as compared with the expected incidence of BC was no longer significant (standardised incidence ratio, 1.15; 95 % CI, 0.86-1.53).

CONCLUSIONS

The multifaceted results of this study warrant further research into the risk and timing of all types of BC recurrence.

Misclassification of second primary and recurrent breast cancer in the surveillance epidemiology and end results registry

The Surveillance Epidemiology and End Results (SEER) registry incorporates laterality, histology, latency, and topography to identify second primary breast cancers. Contralateral tumors are classified as second primaries, but ipsilaterals are subject to additional inclusion criteria that increase specificity but may induce biases. It is important to understand how classification methods affect accuracy of second tumor classification. We collected estrogen, progesterone, and human epidermal growth factor receptor 2 (ER, PR, Her2) status for 11,838 contralateral and 5,371 ipsilateral metachronous secondary tumors and estimated concordance odds ratios (cORs) to evaluate receptor dependence (the tendency for tumors to share receptor status) by laterality. If only second primaries are included, receptor dependence should be similar for contralateral and ipsilateral tumors. Thus, we compared ratios of cORs as a measure of inaccuracy. Cases who met ipsilateral second primary criteria were younger and had less aggressive primary tumor characteristics compared to contralateral tumors. Time to secondary tumors was (by definition) longer for ipsilaterals than contralaterals, especially among ER + primaries. Overall and in multiple strata, ipsilateral tumors showed higher receptor dependence than contralateral tumors (ratios of cORs > 1), suggesting some SEER-included ipsilaterals are recurrences. SEER multiple primary criteria increase specificity, but remain inaccurate and may lack sensitivity. The dearth of early occurring ipsilateral tumors (by definition), coupled with high observed receptor dependence among ipsilaterals, suggests important inaccuracies. Datasets that allow comparison of pathologist- and SEER-classification to true multi-marker genomic dependence are needed to understand inaccuracies induced by SEER definitions.

Correlation Between Disease-Free Survival Endpoints and Overall Survival in Elderly Patients with Early-Stage HER2-Negative Breast Cancer: A SEER-Medicare Analysis

INTRODUCTION

Recent trial-level meta-analyses have established disease-free survival (DFS) as a valid surrogate for overall survival (OS) in human epidermal growth factor receptor 2-negative (HER2-) breast cancer (BC), irrespective of disease stage, and in early-stage hormone receptor-positive (HR+)/HER2- BC. To advance the understanding of the association between additional DFS endpoints and OS, this study assessed the patient-level correlations between DFS and OS, invasive DFS (IDFS) and OS, and distant DFS (DDFS) and OS in Medicare beneficiaries with early-stage HER2- BC, overall and in subgroups of patients with HR+/HER2- BC and triple-negative BC (TNBC).

METHODS

Patients with stages I-III HER2- BC aged ≥ 66 years were identified from SEER-Medicare data (2010-2019). DFS, IDFS, DDFS, and OS were assessed using Kaplan-Meier analyses. Normal scores rank correlation was estimated between each DFS endpoint and OS, overall and separately in patients with HR+/HER2- BC and TNBC.

RESULTS

Of 28,655 patients, 90.4% had HR+/HER2- BC and 9.6% had TNBC (median follow-up 4 years). Median DFS, IDFS, and DDFS were 4.5, 5.9, and 6.3 years, respectively, in HR+/HER2- BC and 3.0, 3.8, and 4.4 years, respectively, in TNBC. Median OS was not reached (5-year OS, HR+/HER2- BC 83.7%; TNBC 67.7%). A significant positive correlation was observed between each DFS endpoint and OS across cohorts, with the strongest correlation observed between DDFS and OS in HR+/HER2- BC (correlation coefficient 0.60; 95% confidence interval 0.57-0.62; p < 0.001) and in TNBC (0.69; 0.65-0.71; p < 0.001).

CONCLUSION

We observed significant positive patient-level correlations between DFS and OS, IDFS and OS, and DDFS and OS in early-stage HER2- BC. Our IDFS and DDFS findings advance the understanding of the role of these DFS endpoints as predictors of OS, and their potential utility as surrogate endpoints in clinical trials of early-stage HER2- BC, given additional validation in trial-level meta-analyses.

Statin use and risks of breast cancer recurrence and mortality

BACKGROUND

Preclinical evidence suggests improved breast cancer survival associated with statin use, but findings from observational studies are conflicting and remain inconclusive. The objective of this study was to assess the association between statin use after cancer diagnosis and cancer outcomes among breast cancer patients.

METHODS

In this retrospective cohort study, 38,858 women aged ≥66 years who were diagnosed with localized and regional stage breast cancer from 2008 through 2017 were identified from the linked Surveillance, Epidemiology, and End Results Medicare database. Statin use was ascertained from Medicare Part D pharmacy claims data. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between post-diagnosis statin use and risks of breast cancer recurrence and breast cancer-specific mortality.

RESULTS

Over a median follow-up of 2.9 years for recurrence and 3.7 years for mortality, 1446 women experienced a recurrence, and 2215 died from breast cancer. The mean duration of post-diagnosis statin use was 2.2 years. Statin use post-diagnosis was not associated with recurrence risk (HR, 1.05; 95% CI, 0.91-1.21), but was associated with a reduced risk of cancer-specific mortality (HR, 0.85; 95% CI, 0.75-0.96). The reduction was more pronounced in women with hormone receptor-positive/human epidermal growth factor receptor 2-negative breast cancer (HR, 0.71; 95% CI, 0.57-0.88).

CONCLUSIONS

These findings suggest that post-diagnosis statin use is associated with improved cancer-specific survival in women with breast cancer and should be confirmed in randomized trials of statin therapy in breast cancer patients.

Estimated incidence of disruptions to event-free survival from non-metastatic cancers in New South Wales, Australia - a population-wide epidemiological study of linked cancer registry and treatment data

Introduction

Population cancer registries record primary cancer incidence, mortality and survival for whole populations, but not more timely outcomes such as cancer recurrence, secondary cancers or other complications that disrupt event-free survival. Nonetheless, indirect evidence may be inferred from treatment data to provide indicators of recurrence and like events, which can facilitate earlier assessment of care outcomes. The present study aims to infer such evidence by applying algorithms to linked cancer registry and treatment data obtained from hospitals and universal health insurance claims applicable to the New South Wales (NSW) population of Australia.

Materials and methods

Primary invasive cancers from the NSW Cancer Registry (NSWCR), diagnosed in 2001-2018 with localized or regionalized summary stage, were linked to treatment data for five common Australian cancers: breast, colon/rectum, lung, prostate, and skin (melanomas). Clinicians specializing in each cancer type provided guidance on expected treatment pathways and departures to indicate remission and subsequent recurrence or other disruptive events. A sample survey of patients and clinicians served to test initial population-wide results. Following consequent refinement of the algorithms, estimates of recurrence and like events were generated. Their plausibility was assessed by their correspondence with expected outcomes by tumor type and summary stage at diagnosis and by their associations with cancer survival.

Results

Kaplan-Meier product limit estimates indicated that 5-year cumulative probabilities of recurrence and other disruptive events were lower, and median times to these events longer, for those staged as localized rather than regionalized. For localized and regionalized cancers respectively, these were: breast - 7% (866 days) and 34% (570 days); colon/rectum - 15% (732 days) and 25% (641 days); lung - 46% (552 days) and 66% (404 days); melanoma - 11% (893 days) and 38% (611 days); and prostate - 14% (742 days) and 39% (478 days). Cases with markers for these events had poorer longer-term survival.

Conclusions

These population-wide estimates of recurrence and like events are approximations only. Absent more direct measures, they nonetheless may inform service planning by indicating population or treatment sub-groups at increased risk of recurrence and like events sooner than waiting for deaths to occur.

Association between pre-diagnosis recreational physical activity and risk of breast cancer recurrence: the California Teachers Study

PURPOSE

Studies have reported inverse associations of pre-diagnosis recreational physical activity (RPA) level with all-cause and breast cancer (BCa)-specific mortality among BCa patients. However, the association between pre-diagnosis RPA level and BCa recurrence is unclear. We investigated the association between pre-diagnosis RPA level and risk of BCa recurrence in the California Teachers Study (CTS).

METHODS

Stage I-IIIb BCa survivors (n = 6,479) were followed with median of 7.4 years, and 474 BCa recurrence cases were identified. Long-term (from high school to age at baseline questionnaire, or, age 55 years, whichever was younger) and baseline (past 3 years reported at baseline questionnaire) pre-diagnosis RPA levels were converted to metabolic equivalent of task-hours per week (MET-hrs/wk). Multivariable Cox proportional hazards models estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for risk of BCa recurrence overall and by estrogen receptor (ER)/progesterone receptor (PR) status.

RESULTS

Long-term RPA was not associated with BCa recurrence risk (ptrend = 0.99). The inverse association between baseline pre-diagnosis RPA level and BCa recurrence risk was marginally significant (≥26.0 vs. <3.4 MET-hrs/wk: HR = 0.79, 95% CI = 0.60-1.03; ptrend = 0.07). However, the association became non-significant after adjusting for post-diagnosis RPA (ptrend = 0.65). An inverse association between baseline pre-diagnosis RPA level and BCa recurrence risk was observed in ER-PR- cases (≥26.0 vs. <3.4 MET-hrs/wk: HR = 0.31, 95% CI = 0.13-0.72; ptrend = 0.04), but not in ER+ or PR+ cases (ptrend = 0.97).

CONCLUSIONS

Our data indicates that the benefit of baseline RPA on BCa recurrence may differ by tumor characteristics. This information may be particularly important for populations at higher risk of ER-PR- BCa.

Association between trajectories of adherence to endocrine therapy and risk of treated breast cancer recurrence among US nonmetastatic breast cancer survivors

BACKGROUND

Endocrine therapy is the mainstay treatment for breast cancer (BC) to reduce BC recurrence risk. During the first year of endocrine therapy use, nearly 30% of BC survivors are nonadherent, which may increase BC recurrence risk. This study is to examine the association between endocrine therapy adherence trajectories and BC recurrence risk in nonmetastatic BC survivors.

METHODS

This retrospective cohort study included Medicare beneficiaries in the United States (US) with incident nonmetastatic BC followed by endocrine therapy initiation in 2010-2019 US Surveillance, Epidemiology, and End Results linked Medicare data. We calculated monthly fill-based proportion of days covered in the first year of endocrine therapy. We applied group-based trajectory models to identify distinct endocrine therapy adherence patterns. After the end of the first-year endocrine therapy trajectory measurement period, we estimated the risk of time to first treated BC recurrence within 4 years using Cox proportional hazards models.

RESULTS

We identified 5 trajectories of adherence to endocrine therapy in BC Stages 0-I subgroup (n = 28,042) and in Stages II-III subgroup (n = 7781). A trajectory of discontinuation before 6 months accounted for 7.0% in Stages 0-I and 5.8% in Stages II-III subgroups, and this trajectory was associated with an increased treated BC recurrence risk compared to nearly perfect adherence (Stages 0-I: adjusted hazard [aHR] = 1.84, 95% CI = 1.46-2.33; Stages II-III: aHR = 1.38, 95% CI = 1.07-1.77).

CONCLUSIONS

Nearly 7% of BC survivors who discontinued before completing 6 months of treatment was associated with an increased treated BC recurrence risk compared to those with nearly perfect adherence among Medicare nonmetastatic BC survivors.

Accuracy of a Cancer Registry Versus Clinical Care Team Chart Abstraction in Identifying Cancer Recurrence

Objective

To evaluate the completeness and reliability of recurrence data from an institutional cancer registry for patients with head and neck cancer.

Patients and Methods

Recurrence information was collected by radiation oncology and otolaryngology researchers. This was compared with the institutional cancer registry for continuous patients treated with radiation therapy for head and neck cancer at a tertiary cancer center. The sensitivity and specificity of institutional cancer registry data was calculated using manual review as the gold standard. False negative recurrences were compared to true positive recurrences to assess for differences in patient characteristics.

Results

A total of 1338 patients who were treated from January 1, 2010, through December 31, 2017, were included in a cancer registry and underwent review. Of them, 375 (30%) had confirmed cancer recurrences, 45 (3%) had concern for recurrence without radiologic or pathologic confirmation, and 31 (2%) had persistent disease. Most confirmed recurrences were distant (37%) or distant plus locoregional (29%), whereas few were local (11%), regional (9%), or locoregional (14%) alone. The cancer registry accuracy was 89.4%, sensitivity 61%, and specificity 99%. Time to recurrence was associated with registry accuracy. True positives had recurrences at a median of 414 days vs 1007 days for false negatives.

Conclusion

Currently, institutional cancer registry recurrence data lacks the required accuracy for implementation into studies without manual confirmation. Longer follow-up of cancer status will likely improve sensitivity. No identified differences in patients accounted for differences in sensitivity. New, ideally automated, data abstraction tools are needed to improve detection of cancer recurrences and minimize manual chart review.

Development of an Automatic Rule-Based Algorithm for the Detection of Ovarian Cancer Recurrence From Electronic Health Records

PURPOSE

As the onset of cancer recurrence is not explicitly recorded in the electronic health record (EHR), a high volume of manual chart review is required to detect the cancer recurrence. This study aims to develop an automatic rule-based algorithm for detecting ovarian cancer (OC) recurrence on the basis of minimally preprocessed EHR data.

METHODS

The automatic rule-based recurrence detection algorithm (Auto-Recur), using notes on image reading (positron emission tomography-computed tomography [PET-CT], CT, magnetic resonance imaging [MRI]), biomarker (CA125), and treatment information (surgery, chemotherapy, radiotherapy), was developed to detect the first OC recurrence. Auto-Recur contains three single algorithms (images, biomarkers, treatments) and hybrid algorithms (combinations of the single algorithms). The performance of Auto-Recur was assessed using sensitivity, specificity, and accuracy of the recurrence time detected. The recurrence-free survival probabilities were estimated and compared with the retrospective chart review results.

RESULTS

The proposed Auto-Recur considerably reduced human resources and time; it saved approximately 1,340 days when scaled to 100,000 patients compared with the conventional retrospective chart review. The hybrid algorithm on the basis of a combination of image, biomarker, and treatment information was the most efficient (sensitivity: 93.4%, specificity: 97.4%) and precisely captured recurrence time (average time error: 8.5 days). The estimated 3-year recurrence-free survival probability (44%) was close to the estimates by the retrospective chart review (45%, log-rank P value = .894).

CONCLUSION

Our rule-based algorithm effectively captured the first OC recurrence from large-scale EHR while closely approximating the recurrence-free survival estimates obtained by conventional retrospective chart reviews. The study findings facilitate large-scale EHR analysis, enhancing clinical research opportunities.

Evaluation of Algorithms Using Automated Health Plan Data to Identify Breast Cancer Recurrences

BACKGROUND

We updated algorithms to identify breast cancer recurrences from administrative data, extending previously developed methods.

METHODS

In this validation study, we evaluated pairs of breast cancer recurrence algorithms (vs. individual algorithms) to identify recurrences. We generated algorithm combinations that categorized discordant algorithm results as no recurrence [High Specificity and PPV (positive predictive value) Combination] or recurrence (High Sensitivity Combination). We compared individual and combined algorithm results to manually abstracted recurrence outcomes from a sample of 600 people with incident stage I-IIIA breast cancer diagnosed between 2004 and 2015. We used Cox regression to evaluate risk factors associated with age- and stage-adjusted recurrence rates using different recurrence definitions, weighted by inverse sampling probabilities.

RESULTS

Among 600 people, we identified 117 recurrences using the High Specificity and PPV Combination, 505 using the High Sensitivity Combination, and 118 using manual abstraction. The High Specificity and PPV Combination had good specificity [98%, 95% confidence interval (CI): 97-99] and PPV (72%, 95% CI: 63-80) but modest sensitivity (64%, 95% CI: 44-80). The High Sensitivity Combination had good sensitivity (80%, 95% CI: 49-94) and specificity (83%, 95% CI: 80-86) but low PPV (29%, 95% CI: 25-34). Recurrence rates using combined algorithms were similar in magnitude for most risk factors.

CONCLUSIONS

By combining algorithms, we identified breast cancer recurrences with greater PPV than individual algorithms, without additional review of discordant records.

IMPACT

Researchers should consider tradeoffs between accuracy and manual chart abstraction resources when using previously developed algorithms. We provided guidance for future studies that use breast cancer recurrence algorithms with or without supplemental manual chart abstraction.

LATTE: Label-efficient incident phenotyping from longitudinal electronic health records

Electronic health record (EHR) data are increasingly used to support real-world evidence studies but are limited by the lack of precise timings of clinical events. Here, we propose a label-efficient incident phenotyping (LATTE) algorithm to accurately annotate the timing of clinical events from longitudinal EHR data. By leveraging the pre-trained semantic embeddings, LATTE selects predictive features and compresses their information into longitudinal visit embeddings through visit attention learning. LATTE models the sequential dependency between the target event and visit embeddings to derive the timings. To improve label efficiency, LATTE constructs longitudinal silver-standard labels from unlabeled patients to perform semi-supervised training. LATTE is evaluated on the onset of type 2 diabetes, heart failure, and relapses of multiple sclerosis. LATTE consistently achieves substantial improvements over benchmark methods while providing high prediction interpretability. The event timings are shown to help discover risk factors of heart failure among patients with rheumatoid arthritis.

Adherence to and persistence with adjuvant hormone therapy, healthcare utilization, and healthcare costs among older women with breast cancer: A population-based longitudinal cohort study

INTRODUCTION

To assess associations between adherence to and persistence with adjuvant hormone therapy, healthcare utilization, and healthcare costs among older women with breast cancer.

MATERIALS AND METHODS

This study was a population-based longitudinal cohort study using the Surveillance, Epidemiology, and End Results (SEER) registry linked with Medicare claims. This study included older women diagnosed with stage I-III hormone receptor-positive breast cancer from 2009 through 2017. Participants were considered adherent with a proportion of days covered (PDC) of 0.80 or more and persistent if they had no hormone therapy discontinuation, i.e., a break of at least 180 continuous days. Length of persistence was calculated as time from therapy initiation to discontinuation. All participants were followed for up to five years after hormone therapy initiation. Generalized linear mixed models with repeated measures or hurdle generalized linear mixed models in the event of excess zeroes were used to assess associations between adherence to and persistence with annual healthcare utilization and costs.

RESULTS

This study included 25,796 women. Being adherent was associated with lower annual healthcare utilization, i.e., hospitalizations, hospital days, emergency room visits, and hospital outpatient visits. Persistence was associated with fewer annual hospitalizations, hospital days, emergency room visits, and hospital outpatient visits. Adherent participants had lower annual inpatient costs, outpatient costs, medical costs, and total healthcare costs despite higher prescription drug costs. Both being persistent and longer persistence were associated with lower inpatient costs, outpatient costs, medical costs, and total healthcare costs despite higher prescription drug costs.

DISCUSSION

This study underscores the economic benefits associated with adherence to and persistence with adjuvant hormone therapy based on comprehensive measures for healthcare utilization and costs. To our best knowledge, this was the first study that reported total healthcare cost savings associated with adherence to and persistence with adjuvant hormone therapy.

Semisupervised Calibration of Risk with Noisy Event Times (SCORNET) using electronic health record data

Leveraging large-scale electronic health record (EHR) data to estimate survival curves for clinical events can enable more powerful risk estimation and comparative effectiveness research. However, use of EHR data is hindered by a lack of direct event time observations. Occurrence times of relevant diagnostic codes or target disease mentions in clinical notes are at best a good approximation of the true disease onset time. On the other hand, extracting precise information on the exact event time requires laborious manual chart review and is sometimes altogether infeasible due to a lack of detailed documentation. Current status labels-binary indicators of phenotype status during follow-up-are significantly more efficient and feasible to compile, enabling more precise survival curve estimation given limited resources. Existing survival analysis methods using current status labels focus almost entirely on supervised estimation, and naive incorporation of unlabeled data into these methods may lead to biased estimates. In this article, we propose Semisupervised Calibration of Risk with Noisy Event Times (SCORNET), which yields a consistent and efficient survival function estimator by leveraging a small set of current status labels and a large set of informative features. In addition to providing theoretical justification of SCORNET, we demonstrate in both simulation and real-world EHR settings that SCORNET achieves efficiency akin to the parametric Weibull regression model, while also exhibiting semi-nonparametric flexibility and relatively low empirical bias in a variety of generative settings.

Survival benefits associated with being adherent and having longer persistence to adjuvant hormone therapy across up to five years among U.S. Medicare population with breast cancer

PURPOSE

To assess associations between adherence to and persistence with adjuvant hormone therapy and mortality among older women with breast cancer.

METHODS

The surveillance, epidemiology, and end results data linked with U.S. Medicare claims was used. This study included older women diagnosed with stage I-III hormone receptor-positive breast cancer from 2009 through 2017. Adherence was defined as having proportion of days covered (PDC) ≥ 0.80. Persistence was defined as having no discontinuation, i.e., no break of ≥ 180 continuous days. Length of persistence was calculated as time from therapy initiation to discontinuation. Cox models with time-dependent covariates were used to assess associations between adherence and persistence with mortality.

RESULTS

This study included 25,796 women. Adherence rates were 78.1 percent, 75.2 percent, 72.4 percent, 70.0 percent, and 61.5 percent from year 1 to year 5 after hormone therapy initiation. Persistence rates were 87.5 percent, 81.7 percent, 77.1 percent, 72.9 percent, and 68.9 percent through cumulative intervals of 1 year up to 5 years. Adherence was associated with all-cause mortality but not associated with breast cancer-specific mortality. Persistent women had lower risk of all-cause mortality and breast cancer-specific mortality. Each additional year of persistence had additional contributions to survival benefits (11% decreased risk of all-cause mortality and 37% decreased risk of breast cancer-specific mortality).

CONCLUSION

This study confirms the detrimental effect of nonadherence to adjuvant hormone therapy across up to 5 years on all-cause survival in older U.S. women. It also reveals the survival benefits associated with having longer persistence across up to 5 years.

Machine Learning Algorithm to Estimate Distant Breast Cancer Recurrence at the Population Level with Administrative Data

Purpose

High-quality population-based cancer recurrence data are scarcely available, mainly due to complexity and cost of registration. For the first time in Belgium, we developed a tool to estimate distant recurrence after a breast cancer diagnosis at the population level, based on real-world cancer registration and administrative data.

Methods

Data on distant cancer recurrence (including progression) from patients diagnosed with breast cancer between 2009-2014 were collected from medical files at 9 Belgian centers to train, test and externally validate an algorithm (i.e., gold standard). Distant recurrence was defined as the occurrence of distant metastases between 120 days and within 10 years after the primary diagnosis, with follow-up until December 31, 2018. Data from the gold standard were linked to population-based data from the Belgian Cancer Registry (BCR) and administrative data sources. Potential features to detect recurrences in administrative data were defined based on expert opinion from breast oncologists, and subsequently selected using bootstrap aggregation. Based on the selected features, classification and regression tree (CART) analysis was performed to construct an algorithm for classifying patients as having a distant recurrence or not.

Results

A total of 2507 patients were included of whom 216 had a distant recurrence in the clinical data set. The performance of the algorithm showed sensitivity of 79.5% (95% CI 68.8-87.8%), positive predictive value (PPV) of 79.5% (95% CI 68.8-87.8%), and accuracy of 96.7% (95% CI 95.4-97.7%). The external validation resulted in a sensitivity of 84.1% (95% CI 74.4-91.3%), PPV of 84.1% (95% CI 74.4-91.3%), and an accuracy of 96.8% (95% CI 95.4-97.9%).

Conclusion

Our algorithm detected distant breast cancer recurrences with an overall good accuracy of 96.8% for patients with breast cancer, as observed in the first multi-centric external validation exercise.

Estimating the incidence of breast cancer recurrence using administrative data

BACKGROUND

Breast cancer is the most common cancer among women, but most cancer registries do not capture recurrences. We estimated the incidence of local, regional, and distant recurrences using administrative data.

METHODS

Patients diagnosed with stage I-III primary breast cancer in Ontario, Canada from 2013 to 2017 were included. Patients were followed until 31/Dec/2021, death, or a new primary cancer diagnosis. We used hospital administrative data (diagnostic and intervention codes) to identify local recurrence, regional recurrence, and distant metastasis after primary diagnosis. We used logistic regression to explore factors associated with developing a distant metastasis.

RESULTS

With a median follow-up 67 months, 5,431/45,857 (11.8%) of patients developed a distant metastasis a median 23 (9, 42) months after diagnosis of the primary tumor. 1086 (2.4%) and 1069 (2.3%) patients developed an isolated regional or a local recurrence, respectively. Patients with distant metastatic disease had a median overall survival of 15.4 months (95% CI 14.4-16.4 months) from the time recurrence/metastasis was identified. In contrast, the median survival for all other patients was not reached. Patients were more likely to develop a distant metastasis if they had more advanced stage, greater comorbidity, and presented with symptoms (p < 0.0001). Trastuzumab halved the risk of recurrence [OR 0.53 (0.45-0.63), p < 0.0001].

CONCLUSION

Distant metastasis is not a rare outcome for patients diagnosed with breast cancer, translating to an annual incidence of 2132 new cases (17.8% of all breast cancer diagnoses). Overall survival remains high for patients with locoregional recurrences, but was poor following a diagnosis of a distant metastasis.

Comparing survival outcomes between neoadjuvant and adjuvant chemotherapy within breast cancer subtypes and stages among older women: a SEER-Medicare analysis

BACKGROUND

This study aimed to compare survival outcomes of neoadjuvant (NAC) and adjuvant chemotherapy (AdC) within each breast cancer subtype and stage among older women.

METHODS

Older (≥ 66 years) women newly diagnosed with stage I-III invasive ductal breast cancer during 2010-2017 and treated with both chemotherapy and surgery within one year were identified from the Surveillance, Epidemiology, and End Results (SEER)-Medicare database. Analyses were performed within each of six groups, jointly defined based on subtype (hormone receptor [HR]-positive/human epidermal growth factor receptor 2 [HER2]-negative, HER2 + , and triple-negative) and stage (I-II and III). Kaplan-Meier curves and multivariable Cox models were used to compare overall and recurrence-free survival between NAC and AdC, with optimal full matching performed for confounding adjustment.

RESULTS

Among 8,495 included patients, 8,329 (20.6% received NAC) remained after matching. Before multiple testing adjustment, Cox models showed that NAC was associated with a lower hazard for death among stage III HER2 + patients (hazard ratio = 0.347, 95% confidence interval CI 0.161-0.745) but a higher hazard for death among triple-negative patients (stage I-II: hazard ratio = 1.558, 95% CI 1.024-2.370; stage III: hazard ratio = 2.453; 95% CI 1.254-4.797). A higher hazard for death/recurrence was associated with NAC among stage I-II HR + /HER2- patients (hazard ratio = 1.305, 95% CI 1.007-1.693). No significant difference remained after multiple testing adjustment.

CONCLUSIONS

The opposite trends (before multiple testing adjustment) of survival comparisons for advanced HER2 + and triple-negative disease warrant further research. Caution is needed due to study limitations such as cancer stage validity.

A semi-supervised adaptive Markov Gaussian embedding process (SAMGEP) for prediction of phenotype event times using the electronic health record

While there exist numerous methods to identify binary phenotypes (i.e. COPD) using electronic health record (EHR) data, few exist to ascertain the timings of phenotype events (i.e. COPD onset or exacerbations). Estimating event times could enable more powerful use of EHR data for longitudinal risk modeling, including survival analysis. Here we introduce Semi-supervised Adaptive Markov Gaussian Embedding Process (SAMGEP), a semi-supervised machine learning algorithm to estimate phenotype event times using EHR data with limited observed labels, which require resource-intensive chart review to obtain. SAMGEP models latent phenotype states as a binary Markov process, and it employs an adaptive weighting strategy to map timestamped EHR features to an embedding function that it models as a state-dependent Gaussian process. SAMGEP's feature weighting achieves meaningful feature selection, and its predictions significantly improve AUCs and F1 scores over existing approaches in diverse simulations and real-world settings. It is particularly adept at predicting cumulative risk and event counting process functions, and is robust to diverse generative model parameters. Moreover, it achieves high accuracy with few (50-100) labels, efficiently leveraging unlabeled EHR data to maximize information gain from costly-to-obtain event time labels. SAMGEP can be used to estimate accurate phenotype state functions for risk modeling research.

Identifying Breast Cancer Recurrence in Administrative Data: Algorithm Development and Validation

Breast cancer recurrence is an important outcome for patients and healthcare systems, but it is not routinely reported in cancer registries. We developed an algorithm to identify patients who experienced recurrence or a second case of primary breast cancer (combined as a “second breast cancer event”) using administrative data from the population of Ontario, Canada. A retrospective cohort study design was used including patients diagnosed with stage 0-III breast cancer in the Ontario Cancer Registry between 1 January 2009 and 31 December 2012 and alive six months post-diagnosis. We applied the algorithm to healthcare utilization data from six months post-diagnosis until death or 31 December 2013, whichever came first. We validated the algorithm’s diagnostic accuracy against a manual patient record review (n = 2245 patients). The algorithm had a sensitivity of 85%, a specificity of 94%, a positive predictive value of 67%, a negative predictive value of 98%, an accuracy of 93%, a kappa value of 71%, and a prevalence-adjusted bias-adjusted kappa value of 85%. The second breast cancer event rate was 16.5% according to the algorithm and 13.0% according to manual review. Our algorithm’s performance was comparable to previously published algorithms and is sufficient for healthcare system monitoring. Administrative data from a population can, therefore, be interpreted using new methods to identify new outcome measures.

Attitudes of Australian breast cancer patients toward the secondary use of administrative and clinical trial data

AIM

Little is known about the attitudes of Australian patients with a history of breast cancer toward the reuse of administrative health data and clinical trial data. Issues of consent, privacy, and information security are key to the discussion. Cancer care and research provides an opportune setting to develop an understanding of attitudes toward data sharing and reuse in individuals with a history of breast cancer.

METHODS

An anonymous, online questionnaire for individuals with a history or diagnosis of breast cancer was distributed by two peak bodies (Breast Cancer Trials [BCT] and Breast Cancer Network of Australia [BCNA]) to their memberships between July 14, 2020 and October 17, 2020. Results were captured in RedCap; data analysis was undertaken using Stata, and a thematic analysis of free text responses was undertaken using NVivo.

RESULTS

One hundred and thirty-two complete responses were received. Twenty-three percent of respondents had participated in a clinical trial, and 12% were currently receiving treatment (chemotherapy, radiotherapy, surgery, or endocrine). Respondents were supportive of the secondary use of de-identified administrative health data and clinical trial data, but showed concern about data security and privacy. Respondents emphasized that the reuse of data should be for improved societal health outcomes, not profit. Many assumed secondary analysis was already undertaken on de-identified administrative health data and clinical trial data.

CONCLUSIONS

Respondents were supportive of the secondary use of de-identified administrative health and clinal trial data within the established bounds of good clinical practice and ethical oversight.

Identifying breast cancer recurrence histories via patient-reported outcomes

PURPOSE

To test accuracy of patient self-report of breast cancer recurrence for enhancing standard population-based cancer registries that do not routinely collect cancer recurrence data despite the importance of this outcome.

METHODS

Potential research subjects were identified in the Breast Cancer Research Database (BCRD) of the Swedish Cancer Institute (SCI). The BCRD has collected data within 45 days of each medical encounter on new primary breast cancer patients receiving all or part of their initial care at SCI. Females diagnosed with a new primary breast cancer 2004-2016, Stages I-III, and alive at the time of study initiation (2018) were identified. Recurrent breast cancer patients were matched 1:1 to surviving non-recurrent patients by patient age, date of diagnosis, and single or multiple primary tumors. Consented research subjects were surveyed about their initial and subsequent diagnostic, therapeutic, and recurrent events. PRO survey responses were compared with BCRD information for each individual participant. Discrepancies were reviewed in medical records.

RESULTS

A matched sample of 88 recurrent and 88 non-recurrent patients were used in analyses. Respondents correctly identified the date of diagnosis of first primary breast cancer within 1 year 94% (165/176). Recurrence was reported by 97% (85/88) of recurrent patients. No recurrence was reported by 100% (88/88) of non-recurrent patients. Recurrence date within 1 year was correctly identified in 79% (67/85). Recurrence site was correctly identified in 82% (70/85). Medical record review of survey-registry discrepancies led to BCRD corrections in 4.5% (8/176) of cases.

IMPLICATIONS FOR CANCER SURVIVORS

Breast cancer patients can accurately report their disease characteristics, treatments, and recurrence history. Patient-reported information would enhance cancer registry data.

New method for determining breast cancer recurrence-free survival using routinely collected real-world health data

Background

In cancer survival analyses using population-based data, researchers face the challenge of ascertaining the timing of recurrence. We previously developed algorithms to identify recurrence of breast cancer. This is a follow-up study to detect the timing of recurrence.

Methods

Health events that signified recurrence and timing were obtained from routinely collected administrative data. The timing of recurrence was estimated by finding the timing of key indicator events using three different algorithms, respectively. For validation, we compared algorithm-estimated timing of recurrence with that obtained from chart-reviewed data. We further compared the results of cox regressions models (modeling recurrence-free survival) based on the algorithms versus chart review.

Results

In total, 598 breast cancer patients were included. 121 (20.2%) had recurrence after a median follow-up of 4 years. Based on the high accuracy algorithm for identifying the presence of recurrence (with 94.2% sensitivity and 79.2% positive predictive value), the majority (64.5%) of the algorithm-estimated recurrence dates fell within 3 months of the corresponding chart review determined recurrence dates. The algorithm estimated and chart-reviewed data generated Kaplan–Meier (K-M) curves and Cox regression results for recurrence-free survival (hazard ratios and P-values) were very similar.

Conclusion

The proposed algorithms for identifying the timing of breast cancer recurrence achieved similar results to the chart review data and were potentially useful in survival analysis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-022-09333-6.

Semiparametric copula-based regression modeling of semi-competing risks data

Semi-competing risks data often arise in medical studies where the terminal event (e.g., death) censors the non-terminal event (e.g., cancer recurrence), but the non-terminal event does not prevent the subsequent occurrence of the terminal event. This article considers regression modeling of semi-competing risks data to assess the covariate effects on the respective non-terminal and terminal event times. We propose a copula-based framework for semi-competing risks regression with time-varying coefficients, where the dependence between the non-terminal and terminal event times is characterized by a copula and the time-varying covariate effects are imposed on two marginal regression models. We develop a two-stage inferential procedure for estimating the association parameter in the copula model and time-varying regression parameters. We evaluate the finite sample performance of the proposed method through simulation studies and illustrate the method through an application to Surveillance, Epidemiology, and End Results-Medicare data for elderly women diagnosed with early-stage breast cancer and initially treated with breast-conserving surgery.

SAT: a Surrogate-Assisted Two-wave case boosting sampling method, with application to EHR-based association studies

OBJECTIVES

Electronic health records (EHRs) enable investigation of the association between phenotypes and risk factors. However, studies solely relying on potentially error-prone EHR-derived phenotypes (ie, surrogates) are subject to bias. Analyses of low prevalence phenotypes may also suffer from poor efficiency. Existing methods typically focus on one of these issues but seldom address both. This study aims to simultaneously address both issues by developing new sampling methods to select an optimal subsample to collect gold standard phenotypes for improving the accuracy of association estimation.

MATERIALS AND METHODS

We develop a surrogate-assisted two-wave (SAT) sampling method, where a surrogate-guided sampling (SGS) procedure and a modified optimal subsampling procedure motivated from A-optimality criterion (OSMAC) are employed sequentially, to select a subsample for outcome validation through manual chart review subject to budget constraints. A model is then fitted based on the subsample with the true phenotypes. Simulation studies and an application to an EHR dataset of breast cancer survivors are conducted to demonstrate the effectiveness of SAT.

RESULTS

We found that the subsample selected with the proposed method contains informative observations that effectively reduce the mean squared error of the resultant estimator of the association.

CONCLUSIONS

The proposed approach can handle the problem brought by the rarity of cases and misclassification of the surrogate in phenotype-absent EHR-based association studies. With a well-behaved surrogate, SAT successfully boosts the case prevalence in the subsample and improves the efficiency of estimation.

Population-based recurrence rates among older women with HR-positive, HER2-negative early breast cancer: Clinical risk factors, frailty status, and differences by race

BACKGROUND

Multiple independent risk factors are associated with the prognosis of hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer (BC), the most common BC subtype. This study describes U.S. population-based recurrence rates among older, resected women with HR+/HER2- early BC.

METHODS

We conducted a retrospective cohort study of older women diagnosed with incident, invasive stages I-III HR+/HER2- BC who underwent surgery to remove the primary tumor using the Surveillance, Epidemiology, and End Results (SEER)-Medicare Linked Database (2007-2015). SEER records and administrative health claims data were used to ascertain patient and tumor-specific characteristics, treatment, and frailty status. Cumulative incidences of BC recurrence were estimated using a validated algorithm for administrative claims data. Multivariable Fine-Gray competing risk models estimated adjusted subdistribution hazards ratios and 95 % confidence intervals for associations with BC recurrence risk.

RESULTS

Overall, 46,027 women age ≥65 years were included in our analysis. Over a median follow up of 7 years, 6531 women experienced BC recurrence with an estimated 3 and 5-year cumulative incidence rates of 10 % and 16 %, respectively. Higher 3- and 5-year cumulative incidences were observed in women with larger tumor size (5+ cm, 21 % and 28 %), lymph node involvement (4+ nodes, 27 % and 37 %), and with frail health status at diagnosis (13 % and 20 %). Independent of these clinical risk factors, Black, Hispanic and American Indian/Alaskan Native women had significantly increased BC recurrence risks.

CONCLUSIONS

Rates of recurrence in HR+/HER2- early BC differs by several patient and clinical factors, including high-risk tumor characteristics. Racial differences in BC outcomes deserve continued attention from clinicians and policymakers.

Development and Use of Natural Language Processing for Identification of Distant Cancer Recurrence and Sites of Distant Recurrence Using Unstructured Electronic Health Record Data

PURPOSE

Large-scale analysis of real-world evidence is often limited to structured data fields that do not contain reliable information on recurrence status and disease sites. In this report, we describe a natural language processing (NLP) framework that uses data from free-text, unstructured reports to classify recurrence status and sites of recurrence for patients with breast and hepatocellular carcinomas (HCC).

METHODS

Using two cohorts of breast cancer and HCC cases, we validated the ability of a previously developed NLP model to distinguish between no recurrence, local recurrence, and distant recurrence, based on clinician notes, radiology reports, and pathology reports compared with manual curation. A second NLP model was trained and validated to identify sites of recurrence. We compared the ability of each NLP model to identify the presence, timing, and site of recurrence, when compared against manual chart review and International Classification of Diseases coding.

RESULTS

A total of 1,273 patients were included in the development and validation of the two models. The NLP model for recurrence detects distant recurrence with an area under the curve of 0.98 (95% CI, 0.96 to 0.99) and 0.95 (95% CI, 0.88 to 0.98) in breast and HCC cohorts, respectively. The mean accuracy of the NLP model for detecting any site of distant recurrence was 0.9 for breast cancer and 0.83 for HCC. The NLP model for recurrence identified a larger proportion of patients with distant recurrence in a breast cancer database (11.1%) compared with International Classification of Diseases coding (2.31%).

CONCLUSION

We developed two NLP models to identify distant cancer recurrence, timing of recurrence, and sites of recurrence based on unstructured electronic health record data. These models can be used to perform large-scale retrospective studies in oncology.

Evaluation of algorithms using administrative health and structured electronic medical record data to determine breast and colorectal cancer recurrence in a Canadian province : Using algorithms to determine breast and colorectal cancer recurrence

BACKGROUND

Algorithms that use administrative health and electronic medical record (EMR) data to determine cancer recurrence have the potential to replace chart reviews. This study evaluated algorithms to determine breast and colorectal cancer recurrence in a Canadian province with a universal health care system.

METHODS

Individuals diagnosed with stage I-III breast or colorectal cancer diagnosed from 2004 to 2012 in Manitoba, Canada were included. Pre-specified and conditional inference tree algorithms using administrative health and structured EMR data were developed. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) correct classification, and scaled Brier scores were measured.

RESULTS

The weighted pre-specified variable algorithm for the breast cancer validation cohort (N = 1181, 167 recurrences) demonstrated 81.1% sensitivity, 93.2% specificity, 61.4% PPV, 97.4% NPV, 91.8% correct classification, and scaled Brier score of 0.21. The weighted conditional inference tree algorithm demonstrated 68.5% sensitivity, 97.0% specificity, 75.4% PPV, 95.8% NPV, 93.6% correct classification, and scaled Brier score of 0.39. The weighted pre-specified variable algorithm for the colorectal validation cohort (N = 693, 136 recurrences) demonstrated 77.7% sensitivity, 92.8% specificity, 70.7% PPV, 94.9% NPV, 90.1% correct classification, and scaled Brier score of 0.33. The conditional inference tree algorithm demonstrated 62.6% sensitivity, 97.8% specificity, 86.4% PPV, 92.2% NPV, 91.4% correct classification, and scaled Brier score of 0.42.

CONCLUSIONS

Algorithms developed in this study using administrative health and structured EMR data to determine breast and colorectal cancer recurrence had moderate sensitivity and PPV, high specificity, NPV, and correct classification, but low accuracy. The accuracy is similar to other algorithms developed to classify recurrence only (i.e., distinguished from second primary) and inferior to algorithms that do not make this distinction. The accuracy of algorithms for determining cancer recurrence only must improve before replacing chart reviews.

Effects of Adjuvant Endocrine Therapy Adherence and Radiation on Recurrence and Survival Among Older Women with Early-Stage Breast Cancer

BACKGROUND

The Cancer and Leukemia Group-B 9343 (CALGB 9343) trial demonstrated that women aged ≥ 70 years with early-stage breast cancer can safely omit radiation therapy (RT) and be treated with breast-conserving surgery (BCS) and adjuvant endocrine therapy (AET) alone. AET adherence is low, leaving an undertreated cohort who may be at increased risk of recurrence and death. We hypothesized that AET adherence and adjuvant treatment choice impact recurrence and survival among CALGB 9343 eligible women.

PATIENTS AND METHODS

SEER-Medicare was used to identify CALGB 9343 eligible women who underwent BCS between 2007 and 2016. Medicare claims were used to identify AET use, and the proportion of days covered by AET was used to categorize adherent (PDC ≥ 0.80) versus nonadherent patients (PDC < 0.80). Recurrence-free, cancer-specific, and overall survival were assessed using Cox proportional hazards models.

RESULTS

In total, 10,719 women were identified, of whom 780 (7.3%) underwent BCS alone, 1490 (13.9%) underwent BCS + RT, 1663 (15.5%) underwent BCS + AET, and 6786 (63.3%) had BCS + RT + AET. Among women treated with BCS + AET, adherent patients had lower recurrence than did nonadherent patients (HR = 0.65, 95% CI: 0.50-0.85). With respect to adjuvant treatment combinations, there was no recurrence difference between the BCS + RT + AET group and BCS + AET group (HR = 0.81, 95% CI: 0.54-1.21). There was equivalent cancer-specific but worse overall survival in the BCS + AET group versus the BCS + AET + RT group.

CONCLUSIONS

While BCS + RT + AET may represent overtreatment for some, AET nonadherent women who omit RT are at risk for worse outcomes. Treatment decisions regarding RT omission should be tailored to the individual patient, taking into consideration the chances of AET nonadherence and the patients' own risk tolerance.

Radiation Without Endocrine Therapy in Older Women With Stage I Estrogen-Receptor-Positive Breast Cancer Is Not Associated With a Higher Risk of Second Breast Cancer Events

PURPOSE

The omission of radiation therapy (RT) in older women with stage 1 estrogen-receptor-positive (ER+) breast cancer receiving endocrine therapy (ET) is an acceptable strategy based on randomized trial data. Less is known about the omission of ET with or without RT.

METHODS AND MATERIALS

We analyzed surveillance, epidemiology, and end results (SEER)-Medicare data for 13,321 women age 66 years or older with stage I ER+ breast cancer from 2007 to 2012 who underwent breast-conserving surgery. Patients were classified into 4 groups: (1) ET + RT (reference); (2) ET alone; (3) RT alone; and (4) neither RT nor ET (NT). Second breast cancer events (SBCEs) were captured using the Chubak high-specificity algorithm. We used χ² tests for descriptive statistics, multivariable multinomial logistic regression to estimate relative risk of undergoing a treatment, and multivariable, propensity-weighted competing-risks survival regression to estimate standardized hazard ratio (SHR) of SBCE. We set significance at P ≤ .01.

RESULTS

Most women underwent both treatments, with 44% undergoing ET + RT, 41% RT alone, 6.6% ET alone, and 8.6% NT, but practice patterns varied over time. From 2007 to 2012, RT decreased from 49% to 30%, whereas ET alone and ET + RT increased (ET alone, 5.4%-9.6%; ET + RT, 38%-51%). Compared with patients age 66 to 69 years, patients age 80 to 85 years were more likely to receive NT (odds ratio [OR], 8.9), RT (OR, 1.9), or ET (OR, 8.8) versus ET + RT (P < .01). Three percent of subjects had an SBCE (2.2% ET + RT, 3.0% RT alone, 3.2% ET alone, 7.0% NT). Relative to ET + RT, NT and ET alone were associated with higher SBCE (NT: SHR, 3.7, P < .001; ET alone: SHR, 2.2, P = .008), whereas RT was not associated with a higher SBCE (SHR 1.21; P = .137). Clinical factors associated with higher SBCE were HER2 positivity and pT1c (SHR, 1.7; P = .006).

CONCLUSIONS

Treatment with RT alone in older women with stage I ER+ disease is decreasing. RT alone is not associated with an increased risk for SBCE. By contrast, NT and ET are both associated with higher SBCE in multivariable analysis with propensity weighting. Further study of the omission of endocrine therapy in this patient population is warranted.

Incidence, survival and stage at diagnosis of small intestinal neuroendocrine tumours in Queensland, Australia, 2001-2015

AIM

Multiple studies have observed increasing incidence of small intestinal (SI) neuroendocrine tumours (NETs). The aim of this study was to describe incidence, mortality and survival of SI NETs by sub-site and stage at diagnosis.

METHODS

Data on patients diagnosed with SI NETs between 2001 and 2015 were sourced from the Queensland Oncology Repository. Staging algorithms utilising several data sources were used to calculate stage at diagnosis (localised, regional or metastatic disease).

RESULTS

We identified 778 SI NETs and of those 716 (92%) had either a documented or derived stage. Incidence doubled from 0.68 per 100 000 to 1.42 per 100 000 over the 15-year period. Most common site was ileum (49.1%) and 84.2% were of carcinoid morphology type. Stage at diagnosis was calculated for 91.7% of patients with 28.3% presenting with regional involvement and 23.9% with distant metastasis. Risk factors associated with metastatic disease were jejunal and SI site not otherwise specified, neuroendocrine carcinoma histology and residing in a rural area. Increasing incidence of localised disease and a corresponding reduction in metastatic disease was observed over time. Five-year cause-specific survival for patients diagnosed between 2001 and 2005 was 82.5%, increasing to 93.8% from 2011 to 2015. Survival was lowest for those with metastatic disease (74.2%). Survival increased between 2001 to 2005 and 2011 to 2015 for each disease stage.

CONCLUSIONS

SI NET incidence in Queensland doubled between 2001 and 2015. Survival was high and improved over time.

Cost-effectiveness of endocrine therapy versus radiotherapy versus combined endocrine and radiotherapy for older women with early-stage breast cancer

PURPOSE

To evaluate the cost-effectiveness of endocrine therapy (ET), radiation therapy (XRT), and combination ET + XRT as post-surgical treatment for older women with early-stage breast cancer from the societal perspective.

METHODS

We constructed a Markov state-transition model consisting of three mutually exclusive health-states: Disease-Free, Recurrence, or Death. Osteoporotic fracture, radiation-induced breast fibrosis, and radiation pneumonitis were modeled as treatment-related adverse events (AEs). Cancer registry-linked-Medicare data were used to assess probability of recurrence and total costs, after propensity adjustment to account for treatment selection, among women aged >65 years diagnosed with estrogen receptor positive or progesterone receptor positive (ER+/PR+) breast cancer receiving ET, XRT, or ET + XRT in 2007-2011. Following randomized controlled trials, overall survival was assumed equivalent, but locoregional recurrence varied. Indirect costs and health-state utilities were literature-driven and varied in sensitivity analyses. Costs and outcomes were discounted at 3% annually.

RESULTS

In a cohort of 10,000 women over ten years, we estimated 1620 total recurrences in the ET-only group, 1296 in the XRT-only group, and 1076 with ET + XRT. Compared to ET-only, the base-case incremental cost-effectiveness ratio (ICER) was $10,826 per quality-adjusted life-year (QALY)-gained for XRT-only and $26,834/QALY-gained for ET + XRT. Similarities in cost and effectiveness between treatments led to highly sensitive results. We also present clinically-relevant patient preference scenarios for recurrence risk-averse patients and near-term AE risk-averse patients.

CONCLUSIONS

The cost-effectiveness of regimens including ET and/or XRT in older women with early-stage breast cancer is sensitive to small differences in costs, as well as risk of, and utilities associated with, locoregional recurrence, suggesting that patient preferences concerning treatment benefits and risks should be considered by physicians.

Adoption and effectiveness of de-escalated radiation and endocrine therapy strategies for older women with low-risk breast cancer

PURPOSE

Recent clinical trials support de-escalation of adjuvant radiation therapy following lumpectomy in some older women with low-risk HR+ breast cancers planning to take endocrine therapy. The adoption of these findings into clinical practice, and the effectiveness of de-escalated therapy in real-world populations, remain under investigation.

MATERIALS AND METHODS

We evaluated use of adjuvant radiation therapy and/or endocrine therapy among older women with T1-2 node-negative, HR+ breast cancer in the United States between 2007 and 2011. The study included patients from the Surveillance, Epidemiology and End Results-Medicare linked database and the North Carolina Cancer Information and Population Health Resource database.

RESULTS

Radiation therapy was received by 65.5% of patients, with no decrease over time. Older women and those with T2 (compared to T1) tumors were less likely to receive radiation therapy. In propensity-adjusted analyses, both radiation therapy alone (HR 0.75, 95% CI 0.67-0.84) and radiation + endocrine therapy (HR 0.62, 95% CI 0.54-0.69) were associated with significantly lower recurrence risk compared to endocrine therapy alone. Non-adherence to endocrine therapy was common (37%) and similar across groups. With a median follow-up of 48 months (range 13-84), we were not able to detect an association of non-adherence with recurrence risk in endocrine therapy-containing treatment arms.

CONCLUSION

Most older women with stage I HR+ breast cancers continue to receive radiation, at higher rates than patients with node-negative stage II tumors. These findings suggest that while multiple evidence-based treatment options exist in these patients, improvements are needed to ensure that radiation therapy is applied equitably and rationally.

Five-Year Adjuvant Endocrine Therapy Adherence Trajectories Among Women With Breast Cancer: A Nationwide French Study Using Administrative Data

BACKGROUND

Adjuvant endocrine therapy (AET) improves long-term survival of breast cancer patients, yet many women are nonadherent or discontinue this treatment. In this study we aimed to describe AET adherence trajectories over 5 years after treatment initiation and to identify factors associated with these trajectories, in a nationwide French cohort of breast cancer survivors.

PATIENTS AND METHODS

Every woman diagnosed with a first nonmetastatic breast cancer in 2011 in France who initiated AET in the 12 months after surgery was included from the French cancer cohort. We identified all reimbursements for AET from national health administrative data sets and modeled AET adherence trajectories over 5 years, using group-based trajectory modeling on the basis of the monthly proportion of days covered by AET. Associated factors were identified using multinomial logistic regressions.

RESULTS

We included 33,260 women. A 6-trajectory model was selected: 1, immediate discontinuation (6.6%); 2, continuous suboptimal adherence (4.3%); 3, progressive nonadherence then discontinuation (6.3%); 4, early nonadherence then discontinuation (5.7%); 5, continuous optimal adherence (68.8%); and 6, late nonadherence then discontinuation (8.3%). The main factors associated with nonadherence trajectories were extreme age (younger than 50 and older than 70 years) and switching AET.

CONCLUSION

Approximately 70% of women had optimal adherence over all 5 years. The original nationwide approach enabled us to identify the "continuous suboptimal adherence trajectory" never previously described.

Classification of drug use patterns

Characterizing long‐term prescription data is challenging due to the time‐varying nature of drug use. Conventional approaches summarize time‐varying data into categorical variables based on simple measures, such as cumulative dose, while ignoring patterns of use. The loss of information can lead to misclassification and biased estimates of the exposure‐outcome association. We introduce a classification method to characterize longitudinal prescription data with an unsupervised machine learning algorithm. We used administrative databases covering virtually all 1.3 million residents of Manitoba and explicitly designed features to describe the average dose, proportion of days covered (PDC), dose change, and dose variability, and clustered the resulting feature space using K‐means clustering. We applied this method to metformin use in diabetes patients. We identified 27,786 metformin users and showed that the feature distributions of their metformin use are stable for varying the lengths of follow‐up and that these distributions have clear interpretations. We found six distinct metformin user groups: patients with intermittent use, decreasing dose, increasing dose, high dose, and two medium dose groups (one with stable dose and one with highly variable use). Patients in the varying and decreasing dose groups had a higher chance of progression of diabetes than other patients. The method presented in this paper allows for characterization of drug use into distinct and clinically relevant groups in a way that cannot be obtained from merely classifying use by quantiles of overall use.

Estimating Population-Based Recurrence Rates of Colorectal Cancer over Time in the United States

BACKGROUND

Population-based metastatic recurrence rates for patients diagnosed with nonmetastatic colorectal cancer cannot be estimated directly from population-based cancer registries because recurrence information is not reported. We derived population-based colorectal cancer recurrence rates using disease-specific survival data based on our understanding of the colorectal cancer recurrence-death process.

METHODS

We used a statistical continuous-time multistate survival model to derive population-based annual colorectal cancer recurrence rates from 6 months to 10 years after colorectal cancer diagnosis using relative survival data from the Surveillance, Epidemiology, and End Results Program. The model was based on the assumption that, after 6 months of diagnosis, all colorectal cancer-related deaths occur only in patients who experience a metastatic recurrence first, and that the annual colorectal cancer-specific death rate among patients with recurrence was the same as in those diagnosed with de novo metastatic disease. We allowed recurrence rates to vary by post-diagnosis time, age, stage, and location for two diagnostic time periods.

RESULTS

In patients diagnosed in 1975-1984, annual recurrence rates 6 months to 5 years after diagnosis ranged from 0.054 to 0.060 in stage II colon cancer, 0.094 to 0.105 in stage II rectal cancer, and 0.146 to 0.177 in stage III colorectal cancer, depending on age. We found a statistically significant decrease in colorectal cancer recurrence among patients diagnosed in 1994-2003 compared with those diagnosed in 1975-1984 for 6 months to 5 years after diagnosis (hazard ratios between 0.43 and 0.70).

CONCLUSIONS

We derived population-based annual recurrence rates for up to 10 years after diagnosis using relative survival data.

IMPACT

Our estimates can be used in decision-analytic models to facilitate analyses of colorectal cancer interventions that are more generalizable.

Validation of an Algorithm to Ascertain Late Breast Cancer Recurrence Using Danish Medical Registries

Purpose

About 70% of women with breast cancer survive at least 10 years after diagnosis. We constructed an algorithm to ascertain late breast cancer recurrence—which we define as breast cancer that recurs 10 years or more after primary diagnosis (excluding contralateral breast cancers)—using Danish nationwide medical registries. We used clinical information recorded in medical records as a reference standard.

Methods

Using the Danish Breast Cancer Group clinical database, we ascertained data on 21,134 women who survived recurrence-free 10 years or more after incident stage I–III breast cancer diagnosed in 1987–2004. We used a combination of Danish registries to construct the algorithm—the Danish National Patient Registry for information on diagnostic, therapeutic and procedural codes; and cancer diagnoses from the Danish Pathology Registry, the Danish Cancer Registry and the Contralateral Breast Cancer database. To estimate the positive predictive value (PPV), we selected 105 patients who, according to our algorithm, had late recurrence diagnosed at Aarhus University Hospital. To estimate the sensitivity, specificity and negative predictive value (NPV), we selected 114 patients diagnosed with primary breast cancer at Aalborg University Hospital. We abstracted clinical information on late recurrence for patients with medical record-confirmed late recurrence at Aarhus University Hospital.

Results

Our algorithm had a PPV of late recurrence of 85.7% (95% CI: 77.5–91.3%), a sensitivity of 100.0% (95% CI, 39.8–100.0%), a specificity of 97.3 (95% CI, 92.2–99.4) and a NPV of 100% (95% CI, 96.6–100.0%).

Conclusion

Our algorithm for late recurrence showed a moderate to high PPV and high sensitivity, specificity and negative predictive value. The algorithm could be an important tool for future studies of late breast cancer recurrence.

A Systematic Review of Estimating Breast Cancer Recurrence at the Population Level With Administrative Data

BACKGROUND

Exact numbers of breast cancer recurrences are currently unknown at the population level, because they are challenging to actively collect. Previously, real-world data such as administrative claims have been used within expert- or data-driven (machine learning) algorithms for estimating cancer recurrence. We present the first systematic review and meta-analysis, to our knowledge, of publications estimating breast cancer recurrence at the population level using algorithms based on administrative data.

METHODS

The systematic literature search followed Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. We evaluated and compared sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy of algorithms. A random-effects meta-analysis was performed using a generalized linear mixed model to obtain a pooled estimate of accuracy.

RESULTS

Seventeen articles met the inclusion criteria. Most articles used information from medical files as the gold standard, defined as any recurrence. Two studies included bone metastases only in the definition of recurrence. Fewer studies used a model-based approach (decision trees or logistic regression) (41.2%) compared with studies using detection rules without specified model (58.8%). The generalized linear mixed model for all recurrence types reported an accuracy of 92.2% (95% confidence interval = 88.4% to 94.8%).

CONCLUSIONS

Publications reporting algorithms for detecting breast cancer recurrence are limited in number and heterogeneous. A thorough analysis of the existing algorithms demonstrated the need for more standardization and validation. The meta-analysis reported a high accuracy overall, which indicates algorithms as promising tools to identify breast cancer recurrence at the population level. The rule-based approach combined with emerging machine learning algorithms could be interesting to explore in the future.

Patterns of Surveillance Advanced Imaging and Serum Tumor Biomarker Testing Following Launch of the Choosing Wisely Initiative

BACKGROUND

The purpose of this study was to assess advanced imaging (bone scan, CT, or PET/CT) and serum tumor biomarker use in asymptomatic breast cancer survivors during the surveillance period.

PATIENTS AND METHODS

Cancer registry records for 2,923 women diagnosed with primary breast cancer in Washington State between January 1, 2007, and December 31, 2014, were linked with claims data from 2 regional commercial insurance plans. Clinical data including demographic and tumor characteristics were collected. Evaluation and management codes from claims data were used to determine advanced imaging and serum tumor biomarker testing during the peridiagnostic and surveillance phases of care. Multivariable logistic regression models were used to identify clinical factors and patterns of peridiagnostic imaging and biomarker testing associated with surveillance advanced imaging.

RESULTS

Of 2,923 eligible women, 16.5% (n=480) underwent surveillance advanced imaging and 31.8% (n=930) received surveillance serum tumor biomarker testing. Compared with women diagnosed before the launch of the Choosing Wisely campaign in 2012, later diagnosis was associated with lower use of surveillance advanced imaging (odds ratio [OR], 0.68; 95% CI, 0.52-0.89). Factors significantly associated with use of surveillance advanced imaging included increasing disease stage (stage III: OR, 3.65; 95% CI, 2.48-5.38), peridiagnostic advanced imaging use (OR, 1.76; 95% CI, 1.33-2.31), and peridiagnostic serum tumor biomarker testing (OR, 1.35; 95% CI, 1.01-1.80).

CONCLUSIONS

Although use of surveillance advanced imaging in asymptomatic breast cancer survivors has declined since the launch of the Choosing Wisely campaign, frequent use of surveillance serum tumor biomarker testing remains prevalent, representing a potential target for further efforts to reduce low-value practices.

Incorporating Breast Cancer Recurrence Events Into Population-Based Cancer Registries Using Medical Claims: Cohort Study

BACKGROUND

There is a need for automated approaches to incorporate information on cancer recurrence events into population-based cancer registries.

OBJECTIVE

The aim of this study is to determine the accuracy of a novel data mining algorithm to extract information from linked registry and medical claims data on the occurrence and timing of second breast cancer events (SBCE).

METHODS

We used supervised data from 3092 stage I and II breast cancer cases (with 394 recurrences), diagnosed between 1993 and 2006 inclusive, of patients at Kaiser Permanente Washington and cases in the Puget Sound Cancer Surveillance System. Our goal was to classify each month after primary treatment as pre- versus post-SBCE. The prediction feature set for a given month consisted of registry variables on disease and patient characteristics related to the primary breast cancer event, as well as features based on monthly counts of diagnosis and procedure codes for the current, prior, and future months. A month was classified as post-SBCE if the predicted probability exceeded a probability threshold (PT); the predicted time of the SBCE was taken to be the month of maximum increase in the predicted probability between adjacent months.

RESULTS

The Kaplan-Meier net probability of SBCE was 0.25 at 14 years. The month-level receiver operating characteristic curve on test data (20% of the data set) had an area under the curve of 0.986. The person-level predictions (at a monthly PT of 0.5) had a sensitivity of 0.89, a specificity of 0.98, a positive predictive value of 0.85, and a negative predictive value of 0.98. The corresponding median difference between the observed and predicted months of recurrence was 0 and the mean difference was 0.04 months.

CONCLUSIONS

Data mining of medical claims holds promise for the streamlining of cancer registry operations to feasibly collect information about second breast cancer events.

Cost-Efficiency of Breast Cancer Care in the US Military Health System: An Economic Evaluation in Direct and Purchased Care

INTRODUCTION

With the rising costs of cancer care, it is critical to evaluate the overall cost-efficiency of care in real-world settings. In the United States, breast cancer accounts for the largest portion of cancer care spending due to high incidence and prevalence. The purpose of this study is to assess the relationship between breast cancer costs in the first 6 months after diagnosis and clinical outcomes by care source (direct or purchased) in the universal-access US Military Health System (MHS).

MATERIALS AND METHODS

We conducted a retrospective analysis of data from the Department of Defense Central Cancer Registry and MHS Data Repository administrative records. The institutional review boards of the Walter Reed National Military Medical Center and the Defense Health Agency reviewed and approved the data linkage. We used the linked data to identify women aged 40-64 who were diagnosed with pathologically-confirmed breast cancer between 2003 and 2007 with at least 1 year of follow-up through December 31, 2008. We identified cancer treatment from administrative data using relevant medical procedure and billing codes and extracted costs paid by the MHS for each claim. Multivariable Cox proportional hazards models estimated hazards ratios (HR) and 95% confidence intervals (CI) for recurrence or all-cause death as a function of breast cancer cost in tertiles.

RESULTS

The median cost per patient (n = 2,490) for cancer care was $16,741 (interquartile range $9,268, $28,742) in the first 6 months after diagnosis. In direct care, women in the highest cost tertile had a lower risk for clinical outcomes compared to women in the lowest cost tertile (HR 0.58, 95% CI 0.35, 0.96). When outcomes were evaluated separately, there was a statistically significant inverse association between higher cost and risk of death (p-trend = 0.025) for women receiving direct care. These associations were not observed among women using purchased care or both care sources.

CONCLUSIONS

In the MHS, higher breast cancer costs in the first 6 months after diagnosis were associated with lower risk for clinical outcomes in direct care, but not in purchased care. Organizational, institutional, and provider-level factors may contribute to the observed differences by care source. Replication of our findings in breast and other tumor sites may have implications for informing cancer care financing and value-based reimbursement policy.

Validation in Alberta of an administrative data algorithm to identify cancer recurrence

Background

Readily available population-based data about cancer recurrence would improve surveillance and research for women of reproductive age.

Methods

We randomly selected 200 women from the Alberta Cancer Registry who had received a cancer diagnosis and who ever had a pregnancy between 2003 and 2012. Administrative data were obtained and linked. Several definitions of recurrence were assessed using various minimum lengths of time between the initial diagnosis date and subsequent diagnoses or treatments, or both. Chart review was used as a "gold standard" definition of recurrence.

Results

Chart review identified recurrences in 26 women. The definition that best captured "recurrence" was 2 or more cancer diagnosis codes 10 or more months from the diagnosis date [sensitivity: 80.8%; 95% confidence interval (ci): 60.7% to 93.5%; specificity: 81.0%; 95% ci: 74.4% to 86.6%; positive predictive value: 38.9%; 95% ci: 25.9% to 53.1%; negative predictive value: 96.6%; 95% ci: 92.2% to 98.9%; kappa = 0.42; 95% ci: 0.28 to 0.57].

Conclusions

Recurrence in reproductive-aged women can be captured with moderate validity using administrative data, but should be interpreted with caution.

Association of Diabetes and Other Clinical and Sociodemographic Factors With Guideline-concordant Breast Cancer Treatment for Breast Cancer

BACKGROUND

Women with breast cancer have worse health outcomes with co-occurring type 2 diabetes, possibly due to suboptimal breast cancer treatment.

METHODS

We created a cohort of women ages 66 to 85 y with stage I to III breast cancer from 1993 to 2012 from an integrated health care delivery system (n=1612) and fee-for-service Medicare beneficiaries (n=98,915), linked to Surveillance, Epidemiology, and End Results (SEER) data (total n=100,527). We evaluated associations between type 2 diabetes and other factors with undergoing guideline-concordant cancer treatment. We estimated χ tests for univariate analysis and relative risks (RRs) using multivariable log-binomial models for outcomes of (1) overall guideline-concordant treatment, (2) definitive surgical therapy (mastectomy or lumpectomy with radiation), (3) chemotherapy if indicated, and (4) endocrine therapy.

RESULTS

Our cohort included 60% of subjects with stage 1 tumors, one quarter below 70 years old, 23% had diabetes, 35% underwent overall guideline-concordant treatment, 24% chemotherapy, and 83% endocrine therapy. Women with diabetes were less likely to undergo overall guideline-concordant treatment (RR: 0.96; 95% confidence interval: 0.94-0.98), and only slightly less likely to undergo guideline-concordant definitive surgical therapy (RR: 0.99; 95% confidence interval: 0.99-1.00). No differences were found for chemotherapy or endocrine therapy. Other factors significantly associated with a lower risk of guideline-concordant care were cancer stages II to III (vs. I; RR=0.47-0.69, P<0.0001), older age (vs. 66 to 69 y; RR=0.56-0.90, P<0.0001), higher comorbidity burden, and Medicaid dual-eligibility.

CONCLUSIONS

Diabetes was associated with lower adherence to overall guideline-concordant breast cancer treatment. However, higher stage, older age, higher comorbidity burden, and Medicaid insurance were more strongly associated with lower use of guideline-concordant treatment. Given the heavy burden of breast cancer and diabetes, long-term outcomes analysis should consider guideline-concordant treatment.

IMPACT

Other factors besides diabetes are more strongly associated with guideline-concordant breast cancer treatment.

The impact of generic aromatase inhibitors on initiation, adherence, and persistence among women with breast cancer: Applying multi-state models to understand the dynamics of adherence

PURPOSE

Clinical trials have clearly documented the survival benefit of aromatase inhibitors (AIs); however, many women fail to initiate (primary nonadherence) or remain adherent to AIs (secondary nonadherence). Prior studies have found that costs impact secondary nonadherence to medications but have failed to examine primary nonadherence. The purpose of this study is to examine primary and secondary adherence following the reduction in copays due to the introduction of generic AIs.

METHODS

Using Surveillance, Epidemiology, and End Results-Medicare data, we identified 50 054 women diagnosed with incident breast cancer between 2008 and 2013. We compare women whose copays would change and those whose would not, due to the receipt of cost-sharing subsidies before and after generics were introduced using a difference-in-difference (DinD) analysis. To examine primary and secondary nonadherence, we rely on a multistate model with four states (Not yet initiated, User, Not Using, and Death). We adjusted for baseline factors using inverse probability treatment weights and then simulated adherence for 36 months following diagnosis.

RESULTS

The generic introduction of AIs resulted in patients initiating AIs faster (DinD = -4.7%, 95%CI = -7.0, -2.3; patients not yet initiating treatment at 6-months), being more adherent (DinD ranging in absolute increase of 8.1%-10.4%) and being less likely to not be using the therapy (DinD range in absolute decrease of 1.2% at 6 months to 8.8% at 24 months) for women that do not receive a subsidy after generics were available.

CONCLUSIONS

Introduction of generic alternatives to AIs significantly reduced primary and secondary nonadherence.

Medical costs associated with metastatic breast cancer in younger, midlife, and older women

PURPOSE

We estimated average medical costs due to metastatic breast cancer (mBC) among younger (aged 18-44), midlife (aged 45-64), and older women (aged 65 and older) by phase of care: initial, continuing, and terminal.

METHODS

We used 2003-2014 North Carolina cancer registry data linked with administrative claims from public and private payers. We developed a claims-based algorithm to identify breast cancer patients who progressed to metastatic disease. We matched breast cancer patients (mBC and earlier stage) to non-cancer patients on age group, county of residence, and insurance plan. Outcomes were average monthly medical expenditures and expected medical expenditures by phase. We used regression to estimate excess costs attributed to mBC as the difference in mean payments between patients with mBC (N = 4806) and patients with each earlier-stage breast cancer (stage 1, stage 2, stage 3, and unknown stage; N = 21,772) and non-cancer controls (N = 109,631) by treatment phase and age group.

RESULTS

Adjusted monthly costs for women with mBC were significantly higher than for women with earlier-stage breast cancer and non-cancer controls for all age groups and treatment phases except the initial treatment among women with stage 3 breast cancer at diagnosis. The largest expected total costs were for women aged 18-44 with mBC during the continuing phase ($209,961 95% Confidence Interval $165,736-254,186).

CONCLUSIONS

We found substantial excess costs for mBC among younger women and during the continuing and terminal phases of survivorship. It is important to assess whether this care is high value for these women.

An augmented estimation procedure for EHR-based association studies accounting for differential misclassification

OBJECTIVES

The ability to identify novel risk factors for health outcomes is a key strength of electronic health record (EHR)-based research. However, the validity of such studies is limited by error in EHR-derived phenotypes. The objective of this study was to develop a novel procedure for reducing bias in estimated associations between risk factors and phenotypes in EHR data.

MATERIALS AND METHODS

The proposed method combines the strengths of a gold-standard phenotype obtained through manual chart review for a small validation set of patients and an automatically-derived phenotype that is available for all patients but is potentially error-prone (hereafter referred to as the algorithm-derived phenotype). An augmented estimator of associations is obtained by optimally combining these 2 phenotypes. We conducted simulation studies to evaluate the performance of the augmented estimator and conducted an analysis of risk factors for second breast cancer events using data on a cohort from Kaiser Permanente Washington.

RESULTS

The proposed method was shown to reduce bias relative to an estimator using only the algorithm-derived phenotype and reduce variance compared to an estimator using only the validation data.

DISCUSSION

Our simulation studies and real data application demonstrate that, compared to the estimator using validation data only, the augmented estimator has lower variance (ie, higher statistical efficiency). Compared to the estimator using error-prone EHR-derived phenotypes, the augmented estimator has smaller bias.

CONCLUSIONS

The proposed estimator can effectively combine an error-prone phenotype with gold-standard data from a limited chart review in order to improve analyses of risk factors using EHR data.

Determining the Time of Cancer Recurrence Using Claims or Electronic Medical Record Data

PURPOSE

Data from claims and electronic medical records (EMRs) are frequently used to identify clinical events (eg, cancer diagnosis, stroke). However, accurately determining the time of clinical events can be challenging, and the methods used to generate time estimates are underdeveloped. We sought to develop an approach to determine the time of a clinical event-cancer recurrence-using high-dimensional longitudinal structured data.

METHODS

Manual chart abstraction provided information regarding the actual time of cancer recurrence. These data were linked to claims from Medicare or structured EMR data from the Cancer Research Network, which were used to determine time of recurrence for patients with lung or colorectal cancer. We analyzed the longitudinal profile of codes that could help determine the time of recurrence, adjusted for systematic differences between code dates and recurrence dates, and integrated time estimates from different codes to empirically derive an optimal algorithm.

RESULTS

We identified twelve code groups that could help determine the time of recurrence. Using claims data for patients with lung cancer, the optimal algorithm consisted of three code groups and provided an average prediction error of 4.8 months. Using EMR data or applying this approach to patients with colorectal cancer yielded similar results.

CONCLUSION

Time estimates were improved by selecting codes not necessarily the same as those used to identify recurrence, combining time estimates from multiple code groups, and adjusting for systematic bias between code dates and recurrence dates. Improving the accuracy of time estimates for clinical events can facilitate research, quality measurement, and process improvement.

Using Medicare Claims to Examine Long-term Prostate Cancer Risk of Finasteride in the Prostate Cancer Prevention Trial

Background

Investigators have used administrative claims to better understand cancer outcomes when a research question cannot feasibly be examined within a study. The Prostate Cancer Prevention Trial (PCPT) showed that seven years of finasteride reduced prostate cancer (PC) risk by 25% in men age 55 years or older. However, it was unclear whether the observed reduction in PC for finasteride participants would be maintained after finasteride discontinuation.

Methods

We examined PC diagnoses identified by PCPT study records and Medicare claims (finasteride = 9423, placebo = 9457). A Medicare-defined PC diagnosis algorithm was defined using diagnosis and procedure codes. Multivariable Cox regression was used to examine time to PC within prespecified follow-up windows (<6.5, 6.5-7.5, and >7.5 years) using time-dependent covariates interacting with intervention assignment to account for the PCPT protocol-specified end-of-study biopsy at seven years. All statistical tests were two-sided.

Results

Median follow-up using the linked database was 16 years. Overall, finasteride arm participants had a 21.1% decrease in the hazard ratio of PC (hazard ratio [HR] = 0.79, 95% confidence interval [CI] = 0.74 to 0.84, P < .001). The beneficial effect of finasteride in reducing the hazard ratio of PC was most pronounced in the first 7.5 years (HR = 0.71, 95% CI = 0.66 to 0.77, P < .001), consistent with the original study findings; after 7.5 years, there was no increased risk of PC for finasteride arm participants (HR = 1.10, 95% CI = 0.96 to 1.26, P = .18).

Conclusions

Finasteride provides a substantial reduction in PC through 16 years of follow-up. There was no strong evidence that the benefit of finasteride diminished after the end-of-study follow-up. Utilizing Medicare claims to augment PCPT follow-up illustrates how the novel use of secondary data sources can enhance the ability to detect long-term outcomes from prospective studies.

Using natural language processing to construct a metastatic breast cancer cohort from linked cancer registry and electronic medical records data

Objectives

Most population-based cancer databases lack information on metastatic recurrence. Electronic medical records (EMR) and cancer registries contain complementary information on cancer diagnosis, treatment and outcome, yet are rarely used synergistically. To construct a cohort of metastatic breast cancer (MBC) patients, we applied natural language processing techniques within a semisupervised machine learning framework to linked EMR-California Cancer Registry (CCR) data.

Materials and Methods

We studied all female patients treated at Stanford Health Care with an incident breast cancer diagnosis from 2000 to 2014. Our database consisted of structured fields and unstructured free-text clinical notes from EMR, linked to CCR, a component of the Surveillance, Epidemiology and End Results Program (SEER). We identified de novo MBC patients from CCR and extracted information on distant recurrences from patient notes in EMR. Furthermore, we trained a regularized logistic regression model for recurrent MBC classification and evaluated its performance on a gold standard set of 146 patients.

Results

There were 11 459 breast cancer patients in total and the median follow-up time was 96.3 months. We identified 1886 MBC patients, 512 (27.1%) of whom were de novo MBC patients and 1374 (72.9%) were recurrent MBC patients. Our final MBC classifier achieved an area under the receiver operating characteristic curve (AUC) of 0.917, with sensitivity 0.861, specificity 0.878, and accuracy 0.870.

Discussion and Conclusion

To enable population-based research on MBC, we developed a framework for retrospective case detection combining EMR and CCR data. Our classifier achieved good AUC, sensitivity, and specificity without expert-labeled examples.

Adherence and Persistence to Adjuvant Hormonal Therapy in Early-Stage Breast Cancer Patients: A Population-Based Retrospective Cohort Study in Israel

Background

Adjuvant hormonal therapy (HT) has been consistently proven to improve multiple outcomes in early breast cancer yet rates of adherence and persistence are variable.

Methods

We retrospectively identified women diagnosed with nonmetastatic breast cancer and initiating HT between January 2000 and December 2007 in a large Israeli health provider. Prescription records including the drug name, date of purchase, and the quantity of pills dispensed were collected. We used Cox proportional hazards and binary logistic models to analyze factors associated with early discontinuation (<5 years) and nonadherence (proportion of days covered, PDC <80%) of HT, respectively.

Results

A total of 4,178 women with breast cancer were identified with nearly 95% of patients treated with tamoxifen as the initial HT. Over the 5-year follow-up period, early discontinuation was identified in 955 (23%) patients. The mean PDC was 82.9% (SD 0.004). Younger age and low BMI were both associated with an increased risk of early discontinuation and nonadherence. A history of hypertension was associated with a higher likelihood of both outcomes.

Conclusion

Adherence and persistence with HT among Israeli breast cancer survivors are comparable to those in international reports. Interventions are necessary to identify and prevent suboptimal HT adherence.