Kaiser Permanente Research Articles

Neighborhood factors associated with late effects among survivors of adolescent and young adult cancer.

The purpose of this study is to evaluate the associations between neighborhood income, education, and neighborhood racial composition (measured as a low percentage of white residents) and risk of developing cardiovascular diseases (CVD), diabetes (DM), and severe depression among survivors of AYA cancer and matched non-cancer peers. Two-year survivors of AYA cancers diagnosed at age 15-39 yrs at Kaiser Permanente Southern California (diagnosed 2000-2012) and individually matched (1:13) non-cancer subjects were included. The development of CVD, DM, and severe depression was ascertained via electronic health records. Neighborhood characteristics were obtained from census-based geocoded data. Cox regression evaluated associations between neighborhood characteristics and the health outcomes of interest among both the cancer survivors and the non-cancer comparison cohort and effect modification by cancer survivor status on these relationships. Among cancer survivors (n = 6774), living in mostly non-white neighborhoods, was associated with risk of CVD (hazard ratio (HR) = 1.54 (95% CI 1.00-2.36)), while lower education level (HR = 1.41 (95% CI 1.02-1.94)) was associated with risk of severe depression. None of the neighborhood characteristics were associated with risk of DM. Effect modification was found for neighborhood education and risk of DM and severe depression. When jointly considered, cancer survivors who resided in the most disadvantaged neighborhoods were at the highest risk of developing these health outcomes compared to other subgroups. Our findings may inform screening strategy and addressing social determinants of health among AYA cancer survivors.

Predicting Self-Reported Social Risk in Medically Complex Adults Using Electronic Health Data.

Social barriers to health care, such as food insecurity, financial distress, and housing instability, may impede effective clinical management for individuals with chronic illness. Systematic strategies are needed to more efficiently identify at-risk individuals who may benefit from proactive outreach by health care systems for screening and referral to available social resources. To create a predictive model to identify a higher likelihood of food insecurity, financial distress, and/or housing instability among adults with multiple chronic medical conditions. We developed and validated a predictive model in adults with 2 or more chronic conditions who were receiving care within Kaiser Permanente Northern California (KPNC) between January 2017 and February 2020. The model was developed to predict the likelihood of a "yes" response to any of 3 validated self-reported survey questions related to current concerns about food insecurity, financial distress, and/or housing instability. External model validation was conducted in a separate cohort of adult non-Medicaid KPNC members aged 35-85 who completed a survey administered to a random sample of health plan members between April and June 2021 (n = 2820). We examined the performance of multiple model iterations by comparing areas under the receiver operating characteristic curves (AUCs). We also assessed algorithmic bias related to race/ethnicity and calculated model performance at defined risk thresholds for screening implementation. Patients in the primary modeling cohort (n = 11,999) had a mean age of 53.8 (±19.3) years, 64.7% were women, and 63.9% were of non-White race/ethnicity. The final, simplified model with 30 predictors (including utilization, diagnosis, behavior, insurance, neighborhood, and pharmacy-based variables) had an AUC of 0.68. The model remained robust within different race/ethnic strata. Our results demonstrated that a predictive model developed using information gleaned from the medical record and from public census tract data can be used to identify patients who may benefit from proactive social needs assessment. Depending on the prevalence of social needs in the target population, different risk output thresholds could be set to optimize positive predictive value for successful outreach. This predictive model-based strategy provides a pathway for prioritizing more intensive social risk outreach and screening efforts to the patients who may be in greatest need.

Intravenous Opioid Administration During Mechanical Ventilation and Use After Hospital Discharge

Guidelines recommend an analgesia-first strategy for sedation during mechanical ventilation, but associations between opioids provided during mechanical ventilation and posthospitalization opioid-related outcomes are unclear. To evaluate associations between an intravenous opioid dose received during mechanical ventilation and postdischarge opioid-related outcomes in medical (nonsurgical) patients. This retrospective cohort study evaluated adults receiving mechanical ventilation lasting 24 hours or more for acute respiratory failure and surviving hospitalization. Participants from 21 Kaiser Permanente Northern California hospitals from January 1, 2012, to December 31, 2019, were included. Data were analyzed from October 1, 2020, to October 31, 2023. Terciles of median daily intravenous fentanyl equivalents during mechanical ventilation. The primary outcome was the first filled opioid prescription in 1 year after discharge. Secondary outcomes included persistent opioid use and opioid-associated complications. Secondary analyses tested for interaction between opioid doses during mechanical ventilation, prior opioid use, and posthospitalization opioid use. Estimates were based on multivariable-adjusted time-to-event analyses, with death as a competing risk, and censored for hospice or palliative care referral, rehospitalization with receipt of opioid, or loss of Kaiser Permanente plan membership. The study included 6746 patients across 21 hospitals (median age, 67 years [IQR, 57-76 years]; 53.0% male). Of the participants, 3114 (46.2%) filled an opioid prescription in the year prior to admission. The median daily fentanyl equivalent during mechanical ventilation was 200 μg (IQR, 40-1000 μg), with terciles of 0 to 67 μg, more than 67 to 700 μg, and more than 700 μg. Compared with patients who did not receive opioids during mechanical ventilation (n = 1013), a higher daily opioid dose was associated with opioid prescriptions in the year after discharge (n = 2942 outcomes; tercile 1: adjusted hazard ratio [AHR], 1.00 [95% CI, 0.85-1.17], tercile 2: AHR, 1.20 [95% CI, 1.03-1.40], and tercile 3: AHR, 1.25 [95% CI, 1.07-1.47]). Higher doses of opioids during mechanical ventilation were also associated with persistent opioid use after hospitalization (n = 1410 outcomes; tercile 3 vs no opioids: odds ratio, 1.44 [95% CI, 1.14-1.83]). No interaction was observed between opioid dose during mechanical ventilation, prior opioid use, and posthospitalization opioid use. In this retrospective cohort study of patients receiving mechanical ventilation, opioids administered during mechanical ventilation were associated with opioid prescriptions following hospital discharge. Additional studies to evaluate risks and benefits of strategies using lower opioid doses are warranted.

INTEGRATING MACHINE LEARNING AND BIG DATA ANALYTICS FOR REAL-TIME DISEASE DETECTION IN SMART HEALTHCARE SYSTEMS

The integration of machine learning (ML) and big data analytics within smart healthcare systems represents a transformative advancement in medical services, emphasizing efficiency, accuracy, and patient-centered care. This paper investigates the application of these advanced technologies in real-time disease detection, showcasing their potential to revolutionize healthcare delivery. Smart healthcare systems leverage a multitude of technological components, including Internet of Things (IoT) devices, sensors, and artificial intelligence (AI), to enable continuous monitoring and diagnostics. This real-time monitoring facilitates prompt interventions and treatment adjustments, which is particularly advantageous for managing chronic conditions and acute illnesses where timely responses are critical to improving patient outcomes. Despite the evident benefits, traditional healthcare infrastructures face significant challenges such as delays in diagnosis due to manual processes, inefficient data handling resulting in data silos, and limited interoperability between different healthcare providers, leading to worsened health outcomes and increased healthcare costs. The integration of ML and big data analytics offers promising solutions to these challenges. ML algorithms can process vast amounts of healthcare data to identify patterns and predict outcomes with high accuracy, such as recognizing early signs of diseases like cancer or diabetes from medical images or electronic health records (EHRs). Big data analytics complements ML by providing the necessary infrastructure to handle and process large volumes of health data, enabling the collection, storage, and analysis of structured data from EHRs, unstructured data from clinical notes, and real-time data from wearable devices. By integrating these technologies, healthcare providers can gain deeper insights into patient health trends and outcomes, leading to more informed decision-making and better patient management. This study employs a qualitative research design, focusing on five genuine case studies: the Mayo Clinic's predictive analytics for heart disease, Cleveland Clinic's use of ML for cancer diagnosis, Kaiser Permanente's diabetes management program, Johns Hopkins Hospital's sepsis detection system, and Mount Sinai Health System's genomic data analysis. Each case study is chosen for its relevance and comprehensive data, detailing the specific healthcare environment and context. This paper interprets these findings in the broader context of smart healthcare systems and existing literature, emphasizing the importance of these technologies in modernizing healthcare and addressing inefficiencies. The challenges encountered during integration, such as data privacy concerns and interoperability issues, are examined along with implemented solutions.

Primary human papillomavirus testing vs cotesting: clinical outcomes in populations with different disease prevalence.

Implementation of primary human papillomavirus (HPV) testing has been slow in the United States perhaps because of concerns of decreased sensitivity compared with concurrent HPV and cytology testing ("cotesting"). We used the National Breast and Cervical Cancer Early Detection Program and the Kaiser Permanente of Northern California cohort to quantify potential trade-offs with primary HPV compared with cotesting in 4 US populations with differing precancer or cancer prevalence. In all settings, cotesting required more lab tests and more colposcopies compared with primary HPV testing. Additional cervical intraepithelial neoplasia grade 3 or cancer immediately detected from cotesting vs primary HPV decreased with decreasing population-average cervical intraepithelial neoplasia grade 3 or cancer prevalence from 71 per 100 000 screened among never or rarely screened individuals in the National Breast and Cervical Cancer Early Detection Program (prevalence = 1212 per 100 000) to 4 per 100 000 screened among individuals with prior HPV-negative results in Kaiser Permanente of Northern California (prevalence = 86 per 100 000). These data suggest that cotesting confer an unfavorable benefit-to-harm ratio over primary HPV testing.

Patient factors associated with chemotherapy modifications: Results from the Optimal Breast Cancer Chemotherapy Dosing (OBCD) study.

520 Background: Understanding patient factors associated with an increased likelihood of receiving chemotherapy modifications is important for treatment planning and anticipating which patient groups might require extra support through the course of treatment. Methods: We have evaluated patient and tumor factors associated with chemotherapy modifications in a cohort of 9700 women receiving adjuvant chemotherapy for breast cancer at Kaiser Permanente Northern California (KPNC) and Kaiser Permanente Washington (KPWA) in the Optimal Breast Cancer Chemotherapy Dosing (OBCD) Study diagnosed between 2005-2019. The outcome variable, receipt of a ‘chemotherapy modification’ is defined as a composite representing any of the following deviations from the intended regimen: dose reduction >10% (first cycle or overall), extensive delay (any single delay >14 days), regimen change, or receipt of fewer cycles of any given drug than intended in the original chemotherapy plan. Multivariable-adjusted prevalence ratios (PR) and corresponding 95% confidence intervals (95% CI) were calculated using generalized linear models of the Poisson family with a log link-function and robust standard errors. Results: Several factors were associated with chemotherapy modifications. Notably, older age was associated with an increased likelihood of modifications (p-trend<0.001), with adults aged 80+ years at diagnosis 1.93-fold more likely to experience modification than adults 18-39 years (95% CI: 1.50-2.50). Increased comorbidity and BMI were also associated with increased likelihood of modification (p-trend for both:<0.001), with patients with a Charlson Comorbidity Index of 3+ (vs 0) having an PR of 1.33 (95% CI: 1.19-1.48) and those with a BMI of 35+ kg/m2 (vs <25 kg/m2) having an PR of 1.53 (95% CI: 1.41-1.65). HER2+ status (PR HER2+ vs HER2-: 1.99 95% CI: 1.89-2.10) and higher stage of disease (PR IIIA vs I:1.24; 95% CI: 1.13-1.35; p-trend<0.001) were associated with increased likelihood of chemotherapy modification, while hormone receptor positivity was associated with lower likelihood of modification (PR ER+ and/or PR+ vs ER-/PR-: 0.93; 95% CI: 0.88-0.99). Diagnosis in more recent years was also associated with decreased likelihood of modification (PR 2015-2019 vs 2004-2009: 0.70; 95% CI: 0.66-0.75), as was higher neighborhood income (PR Q4 vs Q1: 0.78; 95% CI: 0.72-0.84). Conclusions: Several patient-level and clinical factors were associated with chemotherapy modifications, including age, comorbidity, BMI, and ER/PR/HER2 status. Future research might focus on the groups noted to have increased risk of treatment modification to better understand the factors underlying these changes, as well as the impact on patient outcomes.

The impact of HIV and CD4 count stratification on all cause mortality for adenocarcinoma and squamous cell lung cancers in the antiretroviral therapy era: Analysis of National Veterans Affairs and Kaiser Permanente Northern cohorts.

8070 Background: Since the widespread introduction of antiretroviral therapy (ART), people living with HIV (PWH) have experienced substantially decreased morbidity and mortality. However, PWH are at higher risk for lung cancer and have decreased cancer survival rates compared to people without HIV (PWoH). Few studies have evaluated histology-specific lung cancer survival by stage and HIV immune control. Methods: A retrospective cohort study was conducted with newly diagnosed non-small cell lung cancer (NSCLC), including PWH with PWoH, identified between 2008 and 2020 from the US Department of Veterans Affairs and Kaiser Permanente, California. Adjusted hazard ratios for 5-year survival by CD4 count (≥ 500 cells/mm3, <500 cells/mm3, ref: PwOH) were estimated using Cox models stratified by subtype (adenocarcinoma, squamous) and TNM (I, II/III, IV) stage. Results: This cohort consisted of 562 PWH (53% adenocarcinoma, 33% squamous cell) and 91,370 PWoH (52% adenocarcinoma, 33% squamous cell, 15% not specified). The PWH cohort tended to be younger (62 vs. 69 years), male (97% vs. 83%), and non-Hispanic Black race (48% vs. 14%), compared to PWoH. Among PWH, 41% had CD4 ≥500 cells/mm3, 59% had CD4 <500 cells/mm3 . The median survival in years for PWoH, PWH CD4 count ≥500 cell/mm3and <500 cells/mm3 were: 2.4 ±0.10, 2.3 ± 0.17, 2.0±0.17, respectively for adenocarcinoma, and 2.1±0.01, 2.3±0.21, 1.6±0.17 respectively for squamous cell carcinoma. The table shows the adjusted HR for all-cause mortality for Adenocarcinoma and Squamous Cell Carcinoma comparing PWH ≥500 and PWH<500 to PWoH (as reference) stratified by stage. Conclusions: Although it is well known that low CD counts (<200) are associated with poor cancer outcomes, higher CD4 counts (CD<500) at the time of lung cancer diagnosis may still play a role in the treatment outcomes of NSCLC. PWH who have CD4 counts ≥500 appear to have similar survival compared to PWoH, but CD4<500, those with (vs. without) HIV have higher rates of mortality for both adenocarcinoma and squamous cell histologies. Adjusted hazard ratios (HR)* with 95% confidence intervals (CI) for 5-year all-cause-mortality associated with HIV by CD4 count (reference: PWoH), across Adenocarcinoma and Squamous Cell Cancer histology and baseline cancer stage. [Table: see text]

Patient characteristics associated with primary prophylactic granulocyte colony-stimulating factor (G-CSF) use among women treated for early-stage breast cancer.

e12605 Background: The National Comprehensive Cancer Network (NCCN) recommends several chemotherapy regimens for early-stage breast cancer that have elevated risk of neutropenia (>20%). In these regimens, primary prophylactic use of granulocyte colony-stimulating factor (G-CSF) is recommended. Little is known about how patient-level factors relate to receipt of prophylactic G-CSF among women receiving these regimens. Methods: In the Optimal Breast Cancer Chemotherapy Dosing (OBCD) Study, we examined use of G-CSF among women with stage I-IIIA breast cancer between 2005-2019 in Kaiser Permanente Northern California (KPNC) and Kaiser Permanente Washington (KPWA). We evaluated the patient-level factors associated with primary prophylactic G-CSF use (≤7 days before to ≤3 days after first day of chemotherapy) among women receiving chemotherapy regimens with high risk of neutropenia, for which the NCCN guidelines recommend prophylactic G-CSF use. Multivariable-adjusted prevalence ratios (PR) and their corresponding 95% Confidence Intervals (95% CI) were calculated using generalized linear models of the Poisson family with a log link-function and robust standard errors. Results: Overall, while all 3,083 women receiving these regimens ultimately received G-CSF, 2412 (79%) received primary prophylactic G-CSF within the period 7 days prior to 3 days post first cycle. Younger women had higher prevalence of prophylactic G-CSF receipt (82.7% in ages 18-39 years vs. 58.1% in 70+ years). Black women had higher prevalence of receipt compared to white women (82.7% in non-Hispanic Black women vs 76.1% in White women). Women with a higher body mass index (BMI) had lower prevalence of receipt compared to normal weight women (73.3% in 35+ kg/m2 group vs 81.5% in 18.5-<25 kg/m2 group). In multivariable models, older women were less likely to receive prophylactic G-CSF (PR70+ vs 18-39: 0.69; 95% CI: 0.57-0.84; p-trend<0.001), as were those with a higher BMI (PRBMI 35+ vs 18.5-<25: 0.92; 95% CI: 0.85, 0.97; p-trend=0.006), and a later year of diagnosis (PR2016-19 vs 2005-07: 0.81; 95% CI: 0.77, 0.85; p-trend<0.001). Although race/ethnicity was not significantly associated with receipt of prophylactic G-CSF use overall in multivariable models, we observed Non-Hispanic Black women to experience slightly higher likelihood of G-CSF receipt than Non-Hispanic White women (PR: 1.08; 95% CI: 1.02, 1.15). Conclusions: Women of older age and those with higher BMI were less likely to receive primary prophylactic G-CSF, as were women treated in later years. Further research looking at timing of G-CSF use is warranted, as are studies examining provider/facility-level factors in relation to patterns of G-CSF use, G-CSF’s impact on survival outcomes, and the real-world impact of decisions around G-CSF use across age, BMI, and racial/ethnic groups.

Clinician-perceived barriers to cervical cancer screening before and during the COVID-19 pandemic at three US healthcare systems

IntroductionThe COVID-19 pandemic posed serious challenges to cancer screening delivery, including cervical cancer. While the impact of the pandemic on deferred screening has been documented, less is known about how clinicians experienced barriers to screening delivery, and, in particular, the role of pre-pandemic barriers to changes reported during the pandemic. MethodsSurvey of clinicians who performed ≥ 10 cervical cancer screening tests in 2019 from Mass General Brigham, Kaiser Permanente Washington, and Parkland Health, the healthcare systems participating in the Population-based Research to Optimize the Screening Process (PROSPR II) consortium (administered 10/2020–12/2020, response rate 53.7 %). ResultsPrior to the pandemic, clinicians commonly noted barriers to the delivery of cervical cancer screening including lack of staff support (57.6%), interpreters (32.5%), resources to support patients with social barriers to care (61.3%), and discrimination or bias in interactions between staff and patients (31.2%). Clinicians who reported experiencing a given barrier to care before the pandemic were more likely than those who did not experience one to report worsening during the pandemic: lack of staff support (odds ratio 4.70, 95% confidence interval 2.94–7.52); lack of interpreters (8.23, 4.46–15.18); lack of resources to support patients in overcoming social barriers (7.65, 4.41–13.27); and discrimination or bias (6.73, 3.03–14.97). ConclusionsClinicians from three health systems who deliver cervical cancer screening commonly reported barriers to care. Barriers prior to the pandemic were associated with worsening of barriers during the pandemic. Addressing barriers to cervical cancer screening may promote resilience of care delivery during the next public health emergency.

Beyond the pain: Decoding opioid use in advanced cancer ED visits.

e24034 Background: Patients with advanced cancer often turn to the emergency department (ED) for pain treatment, impacting their quality of life and escalating healthcare costs. A crucial step in reducing avoidable ED visits is to identify the characteristics of this patient group and explore potential intervention points. This study aims to characterize the use of narcotics in the ED among patients with advanced cancer. Methods: This retrospective cross-sectional study involved adults aged 18 and older diagnosed with Stage IV solid cancers who passed away within one year of diagnosis (excluding those who died within 30 days). Participants had at least one ED visit within that year at one of 21 cancer centers within Kaiser Permanente Northern California from 2018 to 2020. Demographics, clinical variables, frequencies, and trends in ED visits were collected and compared based on opioid utilization. Pain management details for those receiving opioids during ED visits were also recorded. Results: The study encompassed 2,554 patients with 6,406 ED visits in their final year of life. Sixty-one percent of these patients received opioids during their ED visits, with a median of three visits per patient. In 56% of visits, patients had filled an opioid prescription within the last 30 days. Morphine (39%) and hydromorphone (29%) were the most commonly administered opioids, primarily intravenously (66%). The prevalent cancer subtypes receiving opioids during ED visits were breast, head and neck, and upper GI. Among patients in treatment with chemo, radiation, or both, 65% received opioids during ED visits, a higher rate compared to the 55% of patients not in treatment. For treated patients, chemotherapy was associated with increased ED visits requiring opioid pain management. Median survival did not differ significantly between those who received opioids during ED visits (130 days) and those who did not (123 days). Conclusions: This study emphasizes the dynamic pain needs of this patient population throughout their final year, leading them to seek pain management from the ED when their pain exceeds their home regimen. Factors influencing these needs include cancer subtype and current treatment status, particularly chemotherapy. Further studies are essential to understanding the variations in pain needs over the final year, enabling providers to better recognize and address changes in patients' pain requirements outside the ED.

Broad-spectrum antibiotic use and risk of early-onset colorectal cancer.

e15653 Background: The incidence of early-onset colorectal cancer (eoCRC) diagnosed under age 50 is rising in the United States, but the underlying causes are largely unknown. Use of broad-spectrum antibiotics can induce intense gut dysbiosis, which is strongly implicated in CRC carcinogenesis. Emerging evidence supports a link between broad-spectrum antibiotic use and CRC. We evaluated the associations between oral broad-spectrum antibiotic use and risk of eoCRC. Methods: We conducted a nested case-control study within Kaiser Permanente Southern California (KPSC). Cases were persons aged 15-49 years with invasive colorectal adenocarcinoma diagnosis between 2009-2021; controls were matched at 10:1 ratio on age, sex at birth, and length of KPSC membership. Those with less than 15 years of prior membership at diagnosis date or controls’ index date were excluded to allow evaluation of exposure up to 15 years prior. We extracted all study data, including use of antibiotics by drug type and days of prescriptions, from KSPC’s electronic medical records. Conditional logistic regression was used to evaluate antibiotic use from three etiologic periods: 10-15, 5-10, and 2-5 years prior to diagnosis/index date, adjusting for potential confounders (narrow-spectrum antibiotics use, clinical and lifestyle risk factors, race/ethnicity, body mass index) with a crude p-value < 0.10. Analyses were conducted overall and separately for colon cancer. Results: A total of 310 cases (mean age at diagnosis 44.5 years, 52% female, 33% non-Hispanic White) and 2,976 controls were included. In the 10-15 years prior to diagnosis/index date, 65% of cases and 62% of controls used broad-spectrum antibiotics. Broad-spectrum antibiotic use was not associated with eoCRC overall [OR = 1.15 (95% CI: 0.88-1.49)]. An increased risk was suggested for > 3 months of cumulative use: OR = 1.07 (0.81-1.42) for ≤1 month; 1.26 (0.87-1.84) for 1-3 months; and 1.59 (0.95-2.65) for > 3 months. The association between exposure and colon adenocarcinoma was: OR = 1.26 (0.91-1.72) for any use; OR = 1.17 (0.83-1.64) for ≤1 month; and OR = 1.50 (0.99-2.27) for > 1 month. No association was observed between broad-spectrum antibiotic use within 5-9.9 years or 2-4.9 years prior to diagnosis and risk of eoCRC overall or for the colon site. Conclusions: Our findings suggest a possible association between cumulative longer-term use of broad-spectrum antibiotics and increased risk of eoCRC. The etiologic window for assessing antibiotic exposure appears to be relevant. Larger studies with long-term follow-up are needed to confirm these findings.

Evaluating the utility of routine survivorship exams in breast cancer: Modality of detection of recurrence in a prospective cohort study.

1629 Background: National Comprehensive Cancer Network surveillance guidelines for invasive breast cancer recommend a history and physical exam every 3-12 months for the first 5 years, followed by annually, and mammography every 12 months. However, frequent in-office visits create time toxicity for breast cancer survivors, may have low utility in recurrence detection, and can create false reassurance. This study evaluates the modality of breast cancer recurrence in an integrated health system with access to primary and specialty care. Methods: We investigated mode of detection of recurrent breast cancer in the Pathways Study, a prospective cohort study of adult women diagnosed with and treated for invasive breast cancer in Kaiser Permanente Northern California, with enrollment from 2006 to 2013, and follow-up through December 2021. In this analysis, we included 4,431 women diagnosed at Stages I-III. Cohort members were followed for recurrences, deaths, and other outcomes. Of primary interest was modality of detection of recurrence, categorized as patient-driven symptoms, surveillance mammogram, diagnostic testing (i.e., other surveillance imaging for recurrence such as PET CT or MRI), routine clinical breast exam, or incidental finding. Results: A total of 617 (13.9%) women were identified as having breast cancer recurrence. Patients with recurrence shared similar demographic characteristics to those without recurrence. The recurrence population had median age of 59.2 years, BMI of 28.2 kg/m2, and racial/ethnic composition of 63% white,13% Hispanic,12% Asian, and 10% black. The median time from diagnosis to recurrence was 4 years. Seventy percent of recurrences were found to have distant metastatic disease, of which only 6.7% had concurrent local nodal and/or breast involvement. For those without metastases, 17.7% had recurrence localized in the ipsilateral breast, 2.8% in both the breast and locoregional nodes, and 9.3% only in locoregional nodes. Most recurrences were detected by patient driven symptoms (69.4%). Among the total cases, 8.1% of patients were diagnosed by routine mammogram, 10.9% with diagnostic testing and 4.5% by incidental finding. Recurrences in 6.8% of patients were detected by routine physician exam. Overall survival for those with recurrence was 31.1% and most deaths were attributed to breast cancer (89%). Conclusions: Our study, the largest investigation of breast cancer recurrence modality of detection in a US community oncology setting with comprehensive care, revealed the majority of recurrences were identified through patient self-detection with only 6.8% detected by physician exam. Most recurrences were distant, underscoring the need for alternative screening modalities to detect metastatic recurrences earlier, especially in the context of emerging treatments and improved survival rates for oligometastatic disease.

Diet quality and cardiometabolic risk factors among breast cancer survivors in the Pathways Study.

11039 Background: Women with breast cancer (BC) have higher risk of developing cardiometabolic conditions compared to women without BC. However, the relationship between healthy eating and onset of cardiometabolic conditions in BC survivors remains unknown. We set out to determine if diet quality at BC diagnosis was related to subsequent development of hypertension, diabetes, and dyslipidemia. Methods: This analysis included 3,415 participants from the Pathways Study, a prospective cohort of women diagnosed with invasive BC (stages I-IV) at Kaiser Permanente Northern California (KPNC) between 2005 and 2013. Food frequency questionnaires were administered at or around the time of BC diagnosis and five diet quality indices (DQI) aligned with healthy eating were evaluated: Dietary Approaches to Stop Hypertension (DASH), a healthy plant-based dietary index (hPDI), 2020 Healthy Eating Index (HEI), American Cancer Society nutrition guidelines (ACS), and the alternate Mediterranean Diet (aMED). Incident hypertension, diabetes, and dyslipidemia were ascertained through electronic health records and participants were followed through first indication of these conditions, disenrollment from KPNC, death, or December 31, 2021. Scores were categorized into ascending quartiles of concordance for each DQI, and multivariable adjusted hazard ratios (HR) and 95% confidence intervals (CI) were estimated. Results: There were 554 (16.2%) incident cases of hypertension, 362 (10.6%) of diabetes, and 652 (19.1%) of dyslipidemia over 39,263 person-years of follow-up. Participants in the highest compared to lowest HEI quartile had lower risks of hypertension (HR=0.70, 95% CI 0.53-0.92, Ptrend<0.01), diabetes (HR=0.56, 95% CI 0.40-0.77, Ptrend<0.001), and dyslipidemia (HR=0.76, 95% CI 0.59-0.99, Ptrend=0.03). Participants in the highest vs lowest hPDI quartile had lower risks of hypertension (HR=0.74, 95% CI 0.55-0.98, Ptrend<0.05) and diabetes (HR=0.65, 95% CI 0.46-0.93, Ptrend=0.02), but not dyslipidemia. Women in the highest vs lowest DASH and ACS quartiles had lower risks of diabetes (DASH:HR=0.56, 95% CI 0.39-0.79, Ptrend<0.001; ACS:HR=0.57, 95% CI 0.40-0.79, Ptrend<0.001), but not hypertension or dyslipidemia. No statistically significant differences for aMED were observed between those in the highest vs lowest quartiles. Overall, HRs were similar across DQIs for hypertension and diabetes, except for aMED, which slightly attenuated for diabetes. Conclusions: Diets concordant with HEI may provide the most overall benefit for preventing cardiometabolic conditions after a BC diagnosis. The benefit of diets aligned with hPDI, DASH, ACS, and aMED appear to vary by cardiometabolic condition. Future analyses will examine how these benefits may be modified by BC treatments. Consuming a healthful diet should be recommended to BC survivors for long-term cardiovascular health.

Use of cell-free tumor DNA in early detection of lung cancer.

e20088 Background: USPSTF recommends annual screening for lung cancer with low-dose computed tomography (LDCT) in adults 50-80 years with 20 pack-year smoking history or current/former smoking in the past 15 years. This prospective study evaluated whether we could detect lung cancer at the time of LDCT screening from cell-free tumor DNA using Guardant Health’s LUNAR-2 assay. Methods: Kaiser Permanente Colorado (KPCO) members with upcoming LDCT screenings were invited via email/phone and recruited into two groups, (1) suspicious lung nodules or (2) no lung nodules. We limited enrollment to patients whose LDCT results were classified under the Lung-RADS (version 1.0) categories of 1-3 (recommended follow-up at 12-month intervals) or category 4 (“suspicious”; recommended follow-up at 3-month intervals). We excluded members with lung nodules > 30mm, those with a history of any hematologic malignancy, history of invasive malignancy within the past 5 years, currently pregnant, or known dementia or cognitive impairment. After informed consent, Guardant Health LUNAR test collection kits were mailed to the participant with sample collection instructions provided. Participants were asked to have their blood drawn on the day of their LDCT visit, or no more than 14 days of the scan. We also collected survey and EHR data on smoking history, LDCT, and other associated factors. The KPCO tumor registry confirmed all reported cancers. Results: The study initiated in September 2020 and ended in 08/2022 with 982 individuals consented and blood samples collected. Participants were on average 68 years old, 56% female, 58% former smokers, and 85% had no personal history of cancer. A total of 15 lung cancer cases and 22 other cancer cases were diagnosed. Data analysis is ongoing and expected to be completed early 2024. Conclusions: This study will demonstrate sensitivity and specificity of the LUNAR-2 assay to detect lung cancer relative to standard of care screening modalities in high-risk populations. We will also evaluate whether factors such as comorbidities, pack year history, or medications affect assay accuracy. The impact of this research can inform blood-based early detection of lung cancer.

Breast cancer polygenic risk score and patient survival outcomes in the Pathways study.

10502 Background: Many common, low-penetrance variants have been identified for breast cancer risk through genome-wide association studies, leading to the development of polygenic scores (PGS) to aggregate the effects of individual variants for risk prediction. Much less is known about genetic determinants of breast cancer-related outcomes. We hypothesized that breast cancer patients with high breast cancer PGS are at risk of a second breast-related adverse event, such as recurrence, contralateral breast cancer, or death. Methods: The Pathways Study is a prospective cohort study of breast cancer survivors enrolled shortly after diagnosis in 2006-2013 at Kaiser Permanente Northern California, with ongoing follow-up. Genome-wide genotype data were available from 3,995 patients for calculation of 4 PGS, including PGS313, PGS4k, PGS5k, and PGS6m. Scores were categorized into tertiles to investigate their associations with 7 survival outcomes including recurrence, contralateral second primary breast cancer, other second primary cancer, and death. Hazard ratios (HR) and 95% confidence intervals (CI) were derived from multivariable Cox proportional hazards regression models. Results: The median age at diagnosis was 60 (23.6-94.8) years, and most (68%) self-identified as White. Many women had estrogen receptor (ER)+ (83.4%), stage I and II tumors (89%), with 13.3% HER2+ tumors. The majority (60%) received lumpectomy and some type of adjuvant therapy (44.3% radiotherapy, 47.0% chemotherapy, and 74.6% received hormonal therapy). By December 31, 2021, the median follow-up time was 10.5 (0.2-14.2) years, with 504 recurrences, 419 total second primary cancers, 146 contralateral second primary breast cancers, 237 non-breast second primary cancers, and 762 deaths (352 breast cancer-specific deaths) observed. In Table 1, patients with medium (T2) and high (T3) PGS313 had higher risk of recurrence, but not contralateral breast cancer or other second primary cancer. They also had higher risk of all-cause death, breast cancer-specific death, total breast cancer event, and invasive disease. The other 3 PGS were not associated with survival outcomes examined. Conclusions: Our study reveals that breast cancer patients with higher PGS313 had higher risk of recurrence, total breast cancer and invasive breast cancer events, and death. These findings highlight a potential role of breast cancer PGS for prognosis, warranting further exploration. [Table: see text]

Revisiting the Association of ECOG Performance Status With Clinical Outcomes in Diverse Patients With Cancer.

The ECOG performance status (PS) scale was developed to support national clinical trials, but the degree to which ECOG PS predicts clinical outcomes in patient subgroups outside of clinical trials is relatively unknown. This study examined associations between ECOG PS and adverse outcomes in a diverse community oncology population. In this retrospective cohort study, demographic and clinical characteristics, including the most recent ECOG PS between January 1, 2017, and December 31, 2019, were examined for patients receiving cancer treatment within Kaiser Permanente Northern California (KPNC). Proportional hazard models were used to evaluate the effect of ECOG PS on adverse outcomes. A total of 21,730 patients were identified. Overall, most patients had an ECOG PS of 0 (42.5%) or 1 (42.5%). In multivariable analysis, an ECOG PS of 3 or 4 was associated with higher risk of 30-day emergency department visits (adjusted hazard ratio [aHR], 3.85; 95% CI, 3.47-4.26), 30-day hospitalizations (aHR, 4.70; 95% CI, 4.12-5.36), and 6-month mortality (aHR, 7.34; 95% CI, 6.64-8.11) compared with an ECOG PS of 0. Additionally, we found that upper gastrointestinal and stage IV cancers were associated with a higher risk of adverse outcomes compared with breast and stage I cancers, respectively. When adjusted for ECOG PS, African American race, Asian race, and female sex were associated with a lower risk of mortality than White race and male sex. An ECOG PS of 3 or 4 was more predictive of mortality in younger patients and those with breast cancer (P<.001). ECOG PS and upper gastrointestinal and stage IV cancers were independently associated with increased risk of emergency department visits, hospitalizations, and mortality, whereas African American and Asian race and female sex were associated with decreased risk of mortality. An ECOG PS of 3 or 4 was more predictive of an increased risk of mortality in younger patients and patients with breast cancer. These findings can enhance the use of ECOG PS for clinical decision-making and defining eligibility for clinical trials.

Evaluation of a predictive biomarker for antibody drug conjugates (ADCs).

3140 Background: Several ADCs are FDA-approved for patients with advanced solid tumors; dozens more are in trials. Predictive biomarkers are lacking, with ADC target expression alone performing poorly across ADCs/tumor types. We previously described the development of ADC Treatment Response Scores (ADC-TRS)—a tissue-based multivariate gene-expression test combining individual ADC target expression with proliferation and adhesion—with high correlation of per-ADC/tumor type biomarker frequency (TRS-High) with corresponding clinical trial ORRs for 9 ADCs (PO1-14-04; SABCS 2023). Here we evaluated ADC-TRS in two cohorts of patients treated with approved ADCs. Methods: Adults with advanced solid tumors from an observational trial (NCT03061305) treated with an ADC with TRS results from FFPE tissue (collected prior to 1st ADC treatment) were eligible; HER2+ patients previously treated with trastuzumab or with hormonal therapy treatment between sample collection/first ADC treatment were excluded. Kaiser Permanente Northern California (KPNC; limited to breast [BR] and bladder [BL] cancer) patients and non-KPNC patient (pan-solid tumor) cohorts were analyzed separately. We evaluated whether treatment-matched ADC-TRS status (High or Low) was significantly associated with overall survival (OS) after ADC treatment initiation by Cox proportional hazards models (adjusting for age, indication, and years since tissue collection). Additional analyses in the non-KPNC cohort evaluated whether proliferation/adhesion-related gene expression provided information on outcomes beyond ADC target expression only and predictive/prognostic effect. Results: Patients treated with an approved ADC were included in the non-KPNC (n=72; 5 ADCs/9 tumor types); most frequent sacituzumab govitecan/BR [43%]) and KPNC (n=127; 3 ADCs/two tumor types; most frequent enfortumab vedotin/BL[50%]) cohorts, respectively. In both cohorts, ADC TRS status was significantly associated with OS (High vs. Low, median OS 22.8 vs. 8.7 months [mo.], aHR 0.15, p=0.005, and median OS 15.3 vs. 8.3 mo., aHR 0.58, p=0.011, respectively). In the non-KP cohort, ADC-TRS significantly improved model fit for OS beyond target gene expression alone (likelihood ratio test [LRT] p=0.0008). A predictive biomarker effect was confirmed by lack of ADC-TRS status association with OS from systemic chemotherapy start in indication-matched patients not treated with an ADC (n=126, aHR 0.43-2.79, p=0.11-1.0 for the 9 individual TRS). In a 15,108 patient pan-tumor cohort (regardless of treatment), 26.7% were ADC-TRS High for at least 1 approved ADC outside approved indications. Conclusions: Results support ADC-TRS as a multivariate gene-expression-based biomarker that predicts ADC OS across tumor types and targets. More than 25% of all patients with advanced solid tumors have a greater likelihood of being responsive to one or more approved ADCs outside of approved indications.

Artificial intelligence–aided lung cancer screening in routine clinical practice: A pilot of LungFlag at Geisinger.

e13604 Background: Use of low-dose computed tomography (LDCT) in lung cancer screening is recommended for early detection and can reduce mortality. LungFlag (Medial-EarlySign) is a machine-learning risk prediction model that identifies individuals at high risk of non-small cell lung cancer so their care team may recommend LDCT screening. The LungFlag model was developed using data from Kaiser Permanente Southern California and is being implemented in a pilot at Geisinger. Here we provide an overview of the clinical workflow for the LungFlag pilot. Methods: Geisinger has implemented LungFlag into routine clinical practice via integration with the electronic medical record system. This pilot is examining the performance and clinical impact of LungFlag in the first year. LungFlag runs every month to provide risk evaluation for individuals 50 to 80 years old who are eligible for annual lung cancer screening according to US Preventive Services Task Force recommendations. Eligible individuals have a 20 pack-year smoking history, no lung cancer history, and currently smoke or quit within the past 15 years. Patients with cancer history in the past 5 years and those screened or scheduled for screening in the past year are excluded. LungFlag identifies the top 30 individuals with the highest LungFlag scores per month (high-risk patients) and refers them for targeted outreach from a lung cancer screening program nurse navigator. A positive LungFlag result prompts nurse navigator chart review, patient outreach, and coordination with the primary care provider. Flagged individuals are contacted (up to 3 phone calls) and encouraged to receive lung cancer screening. Individuals not flagged by the model (non–high-risk patients) receive usual care consisting of reminders for lung cancer screening during primary care visits without additional targeted outreach and high-risk prioritization. Results: The LungFlag pilot was initiated in August 2023 and will continue until August 2024. As of January 2024, risk profiles of &gt;12,000 patients were analyzed and 180 high-risk patients were identified. Conclusions: This pilot is the first evaluation of the implementation process for the LungFlag model in routine clinical practice. After the pilot, a retrospective observational study is planned to further assess the impact of LungFlag and the associated clinical workflow in identifying high-risk individuals and increasing compliance with LDCT screening among high-risk individuals.

Treatment rates and reasons for non-treatment in a real-world cohort of patients with pancreatic cancer.

e13554 Background: The proportion of patients who receive treatment for pancreatic ductal adenocarcinoma (PDAC) and reasons for non-treatment in real-world patient populations are poorly understood. Our study aims to determine treatment rates, predictors for non-treatment, and the rationale for declining treatment in a well-defined, diverse, community-based patient population. Methods: We identified all adult patients with PDAC (n = 3852) between 2010 and 2020 through the Kaiser Permanente Northern California Cancer Registry. Patients were stratified by age and tumor characteristics to determine rates of oncologic therapy. Multivariable logistic regression was used to assess how covariates were associated with receiving treatment. A detailed chart review of all patients ≤ 65 years and a random sample of 100 patients &gt; 65 years with non-metastatic disease who did not receive treatment was performed to better understand the reasons for non-treatment. A random sample of 20 patients with metastatic disease who did not receive treatment were also reviewed. Results: Of patients with PDAC, 77.7% of patients with non-metastatic and 56.2% with metastatic disease received any cancer-directed treatment. A multivariate analysis showed no statistically significant differences by race/ethnicity or gender. Increasing age was a significant negative predictor of treatment. Charlson comorbidity index (CCI) did not influence treatment rates in patients with non-metastatic PDAC, but in patients with metastatic disease those with CCI 3+ had a decreased odds of receiving treatment. Of patients ≤ 65 years with non-metastatic disease, 90.3% were seen by an oncologist, and of those patients 89.3% were offered treatment. The primary reason for declining treatment was patient/family preference in 80.8%, followed by PDAC-related frailty in 11.5%, and medical comorbidity/other malignancy in 7.69%. Similarly, of patients &gt; 65 years with non-metastatic disease, the majority were also seen by an oncologist, offered treatment, and declined due to personal preference. The same trends were noted in patients with metastatic PDAC. Conclusions: A significant proportion of patients with PDAC did not receive cancer-directed treatment in a real-world, community-based patient population, and this was more likely for patients with metastatic disease. Race and gender did not influence whether patients received treatment in this fully insured patient population and comorbidity had a small effect. Patients of increasing age, however, were less likely to receive treatment. Most patients who did not initiate treatment did so primarily due to patient/family preference despite being offered cancer-directed therapy by their treatment team. These findings may help guide the development and utilization of patient decision-aids in PDAC to improve goal concordant care that accounts for patients’ preferences and values.

Real-world treatment patterns and outcomes of zanubrutinib in chronic lymphocytic leukemia and small lymphocytic leukemia (CLL/SLL).

11158 Background: Zanubrutinib (zanu) is a next-generation selective Bruton’s tyrosine kinase inhibitor (BTKi) with superior efficacy over first generation ibrutinib (ibru) in CLL/SLL patients (pts). Herein we present real-world treatment patterns based on a formulary change from ibrutinib to zanubrutinib in pts with CLL/SLL in an integrated community oncology practice. Methods: We retrospectively analyzed CLL/SLL pts 18 years and older who received at least 3 months of zanubrutinib from October 1, 2018 to September 15, 2023 at Kaiser Permanente Northern California. Treatment patterns, treatment-emergent adverse events (TEAEs: AEs reported during BTKi use), treatment-limiting adverse events (TLAEs: AEs leading to BTKi discontinuation), and mortality were reported. Results: A total of 281 pts received zanu (median age: 71 years; 64% male); 190 pts switched from ibru (ibru-zanu), and 91 pts received zanu only. Most pts were White (75%) followed by Black (10%). Compared with ibru-zanu pts, zanu-only pts were older (median age 74 vs 69 years) and had more comorbidities but were comparable in sex, race, and insurance type. The primary reasons for switching to zanu were formulary change (73%) and progression (15%). Median follow-up time after initiation of first BTKi was longer in the ibru-zanu group (Table). Similar TEAE rates were seen with use of both BTKi therapies, with lower TLAE rates with zanu (Table). Most common TLAE were atrial fibrillation and fatigue for ibru, and cytopenias and rash/bruising for zanu. Cardiac TLAE and non-TLAE rates overall were higher with ibru than zanu, and the rates decreased while on zanu after switching from ibru (Table). Dose modification occurred in 34 pts on ibru and 50 pts on zanu (18 ibru-zanu pts, 32 zanu-only pts), with reductions primarily for grade 1-2 AEs. Of the 281 pts who received zanu, 79% remain on treatment at the end of data collection; 13 pts died (8 from infection, including 5 from COVID), with no reports of treatment-related deaths. Conclusions: In the real-world setting post-formulary change, zanu is effective and safe in pts with or without prior ibru use. Zanu use had lower cardiotoxicity and TLAE rates than ibru though data was limited by a difference in follow-up time. Similar results were seen in zanu-only pts despite being older and having more comorbidities, with discontinuation most often due to grade &lt;3 AEs. [Table: see text]