Cary Philip Gross

Introduction: While broad genomic profiling is increasingly used in advanced non-small cell lung cancer (aNSCLC), the impact of test results on subsequent guideline-concordant targeted therapy selection remains incompletely understood. Methods: Using a merged dataset of 2 large, nationwide, patient-level databases, we identified patients who were diagnosed with aNSCLC between 2017-2023, had potentially actionable genomic profiling findings, and initiated systemic therapy. Patients were categorized into actionability subgroups based on contemporaneous regulatory approvals and NCCN guideline recommendations. Within each subgroup, we assessed receipt of guideline-concordant targeted therapy within 24 months, including potential underuse (non-receipt of recommended treatment) and overuse (receipt of non-recommended treatment). Results: Among 6620 patients (67.4% ≥65 years, 54.6% female, 68.9% White), guideline-concordant targeted therapy use varied substantially by actionability category: 2313 (89.6%) of 2582 patients with available first-line on-label options received them (10.4% underuse), while 212 (67.3%) of 315 patients with available later-line on-label options received them after first-line (32.7% underuse). Among 441 patients with available guideline-concordant off-label options, only 122 (27.7%) received them (72.3% underuse). Conversely, 238 (8.6%) of 3282 patients received matched but guideline-discordant off-label options, representing overuse of ineffective or unestablished therapies. Smoking history, squamous histology, and high PD-L1 expression were associated with lower targeted therapy receipt. Conclusions: In this cohort study of aNSCLC care, the guideline concordance of targeted therapy use varied by clinical actionability of molecular testing results. Underuse was more common in patients with later-line and off-label targeted therapy options. Patients with classical smoking-related risk profiles were substantially less likely to receive targeted therapy even when actionable alterations were identified.

Electronic health record (EHR)-linked cohorts support association, prediction, and causal studies using longitudinally measured markers of health. However, a lab biomarker measurement is recorded only when a patient first has a medical encounter (visit process) and, a clinician orders the corresponding test and the patient follows through (observation process). These two stages may induce informative presence (IP) and informative observation (IO), respectively. Yet their drivers remain largely uncharacterized, despite evidence that understanding this recording mechanism is essential for selecting appropriate strategies for downstream analysis that treat these markers as longitudinally measured outcomes. We characterize this two-stage recording hierarchy using a stochastic recurrent-event model for the outpatient visit process and a visit-process-weighted generalized estimating equation model for biomarker recording conditional on an outpatient visit. We characterize descriptors of both processes in three EHR-linked cohorts in the US (All of Us [AoU], n=599,423; Yale New Haven Health System [YNHHS], n=319,666; Michigan Genomics Initiative [MGI], n=82,372), reporting descriptive statistics for longitudinal visits and for a panel of 68 lab biomarkers commonly measured in EHRs. We conduct detailed model-based analyses of ten biomarkers spanning multiple domains: routine monitoring, general laboratory assessment, and symptom-triggered testing. These include glucose, hemoglobin A1c [HbA1c], creatinine, hemoglobin [Hgb], white blood cell count [WBC], low-density lipoprotein [LDL] and high-density lipoprotein [HDL] cholesterol, triglycerides, C-reactive protein [CRP], and thyroid-stimulating hormone [TSH]. Across the three cohorts, the median number of outpatient visits ranged from 1.7 to 6.1 per year over a median follow-up of 4.4 to 7.2 years. Among patients with at least one recorded measurement, the median within-person proportion of visits containing a given biomarker ranged from 0.4% to 19.5%, demonstrating that more frequent visits did not necessarily translate into greater per-visit biomarker capture. In the visit-process models, chronic disease burden, and a recent history of outpatient visits were consistently associated with higher visit rates across all three cohorts whereas associations with race, ethnicity, and neighborhood-level income varied across cohorts. In per-visit observation models, the association of covariates depended on the biomarker under consideration; for example, prior cancer diagnosis was associated with more frequent measurement of blood counts but with less frequent measurement of lipids. These findings provide a deeper understanding of how to model who seeks care and what is measured as two distinct recording processes in EHR. Our empirical findings show that the descriptors of these processes vary across cohorts and biomarkers, providing guidance on how to construct these models for downstream longitudinal analyses with irregular EHR visits.

Background: Understanding the prevalence of financial conflicts of interest is important in United States oncology, where cancer burden and annual expenditures are high. We estimated the financial conflicts of interest prevalence across the US cancer ecosystem: (1) clinical trials, (2) US Food and Drug Administration meeting participants, (3) academic medical leaders, (4) patient advocacy organizations, (5) authors of clinical practice guidelines and pathways, and (6) practicing oncologists. Methods: We searched Ovid MEDLINE, Ovid Embase, Scopus, Web of Science, and the Healthcare Administration Database for oncology primary research studies and systematic reviews published since 2015 reporting a percentage of industry-related financial conflicts of interest. Two reviewers (C.K., L.W.) extracted data from eligible studies using an explicit extraction form and reached consensus. We report narrative summaries of each domain and conduct random effects meta-analysis of financial conflicts of interest prevalence on domains with at least 5 eligible studies. Results: Among 36 eligible articles, 30 reported more than 50% prevalence of financial conflicts of interest in the sample. Financial conflicts of interest prevalence ranges varied within domains: 36.8%-89.0% of cancer clinical trials, 65%-92.1% of US Food and Drug Administration meeting speakers, 76%-100% among academic medical leaders, 56%-98.1% of patient advocacy organizations, 46.4%-98.4% of clinical practice guideline and pathway authors, and 24.7%-100% of clinical oncologists. Two domains (cancer trials and practicing oncologists) were eligible for meta-analysis. The pooled prevalence of industry sponsorship in clinical trials was 69% (95% CI = 62% to 77%), and the pooled prevalence of industry general payments to practicing oncologists was 61% (95% CI = 47% to 75%). Conclusion: Financial conflicts of interest prevalence exceeds 50% across multiple levels of the cancer ecosystem, warranting close monitoring to limit impact on clinical decision-making and patient outcomes.

Importance: Broad genomic profiling (BGP) is recommended for several types of metastatic cancer but remains underused. Over half of Medicare beneficiaries are enrolled in Medicare Advantage (MA), where cost-containment strategies may limit access to BGP. Whether Medicare payer type and geographic region are associated with BGP use is not well established. Objectives: To evaluate whether BGP use differs by Medicare payer type (MA vs fee-for-service Medicare [FFS]) and to characterize geographic variation in BGP use across hospital referral regions (HRRs). Design, Setting, and Participants: This nationwide retrospective cohort study used Medicare Chronic Conditions Data Warehouse claims and service records to identify beneficiaries aged 66 years or older with a new diagnosis of metastatic cancer, including bladder, breast, colorectal, endometrial, kidney, lung, melanoma, pancreatic, prostate, or thyroid, from January 1, 2020, to June 30, 2022. Data analysis was conducted from October 2024 to March 2026. Exposures: Medicare type (FFS vs MA) and HRR. Main Outcomes and Measures: The primary outcome was receipt of BGP within 2 months before through 6 months after diagnosis. Mixed-effects logistic regression models were used to estimate adjusted odds ratios (AORs) for the association between Medicare type and BGP use, controlling for demographic, clinical, and geographic factors. HRR-level variation was summarized using the median odds ratio (MOR). Subgroup analyses stratified cancers by the strength of guideline recommendations for BGP. Results: Of 254 720 Medicare beneficiaries with metastatic cancer (median age, 74 years [IQR, 70-79 years]; 141 964 female [55.7%]), 112 637 (44.2%) were enrolled in MA and 142 083 (55.8%) in FFS. Overall, 64 351 (25.3%) received BGP. FFS beneficiaries had higher BGP use than MA beneficiaries (36 633 of 142 083 [25.8%] vs 27 718 of 112 637 [24.6%]; AOR, 1.08 [95% CI, 1.06-1.10]). BGP use was more frequent among FFS vs MA beneficiaries for cancers with equivocal BGP recommendations (AOR, 1.15 [95% CI, 1.11-1.19]) and, to a lesser extent, cancers with explicit recommendations (AOR, 1.04 [95% CI, 1.02-1.07]). Adjusted BGP use varied widely across HRRs (range, 13.8%-35.9%; median, 24.5% [IQR, 21.8%-27.6%]; MOR, 1.28 [95% CI, 1.25-1.31]). Conclusions and Relevance: In this cohort study of Medicare beneficiaries with metastatic cancer, BGP use differed by Medicare payer type and showed substantial regional variation. These findings highlight opportunities to improve guideline-concordant molecular testing.

Background: Survival disparities in prostate cancer are driven in part by social and economic factors, including housing insecurity. Rent subsidies in the form of federal housing assistance are a well-established strategy for alleviating housing insecurity, but their association with prostate cancer care and survival is unknown. Methods: Using linked federal housing assistance data, SEER cancer registry data, and Medicare claims, we assessed workup and treatment receipt and two-year survival among individuals aged 66 to 95 who were diagnosed with prostate cancer in 2007 to 2019. We used logistic and Cox regression models to compare outcomes between individuals receiving housing assistance upon prostate cancer diagnosis and a comparison group without housing assistance, using propensity score matching to balance sociodemographic and clinical characteristics. Results: There were 1,839 individuals with housing assistance (and 5,517 without assistance) included in the workup and treatment analyses and 4,451 individuals with housing assistance (and 13,353 without assistance) in the survival models. Receipt of housing assistance was not associated with guideline-concordant workup (62.7% vs 61.2%, OR 1.02 [95% CI 0.97, 1.07], p = 0.48) or active treatment (63.7% vs 62.2%, OR 1.02 [95% CI: 0.99, 1.06], p = 0.22) but was associated with improved overall survival in the 2 years following diagnosis (hazard ratio for mortality: 0.88 [95% CI: 0.81, 0.96]). Conclusions: Older adults receiving housing assistance at the time of a prostate cancer diagnosis experienced better overall survival than a matched comparison group without housing assistance. Results suggest that expanding access to housing assistance might support greater and more equitable survival in prostate cancer.

Background: Health-related social conditions (HRSCs) are modifiable factors affecting ovarian cancer outcomes. How best to manage HRSCs as part of ovarian cancer care remains unclear. We sought interest-holder perspectives on HRSC assessment and assistance to inform integration of social care with gynecologic oncology practice. Methods: In this qualitative descriptive study, we conducted individual, semi-structured interviews with patients with ovarian cancer, their caregivers, and oncology clinicians from an academic medical center. We solicited thoughts on HRSC assessment and assistance, including whether and how HRSC discussions should occur, as well as opinions on a 20-item HRSC assessment and a community resource referral document. We analyzed data using the Rigorous and Accelerated Data Reduction (RADaR) technique. Results: Participants (n = 14) included 5 patients with ovarian cancer, 4 caregivers, and 5 clinicians. Patients and caregivers were open to HRSC conversations with clinicians from various disciplines if undertaken with empathy, respect, and active listening. In addition to common HRSCs like food insecurity and housing instability, participants endorsed attending to religion/spirituality and assessing for financial hardship, interpersonal violence, psychological support, and neighborhood safety. Patients and clinicians characterized the HRSC screening questions as straightforward and in-depth. All participants felt the community resource referral document was well-organized and helpful. Conclusion: Even in the context of a life-threatening disease, participants affected by and caring for patients with ovarian cancer regard social conditions as relevant to cancer care. Our findings inform practical considerations of who, what, when, where, why, and how HRSC assessment and assistance can be undertaken in oncology.

Background/Aims: Clinical trials are commonly believed to benefit from the involvement of an academic principal investigator who accepts responsibility for design, conduct, and reporting. Little evidence exists, however, about the importance that diverse stakeholders assign to the principal investigator’s role in leadership of trials. Furthermore, few studies have examined whether and how beliefs about the role of the principal investigator might vary by funding source. Methods: We conducted parallel Delphi panel surveys with seven stakeholder groups (principal investigators, patient advocates, journal editors, public funders, industry representatives, United States Food and Drug Administration officials, and clinical trial cooperative-group chairs) to assess the extent to which respondents believed leadership of a multicenter randomized controlled trial by an academic principal investigator to be important, considering publicly and industry-funded trials separately. We then surveyed an international sample of principal investigators (N = 92) who had recently published a multicenter randomized controlled trial in a high-impact general medical, oncology, cardiovascular, or psychiatry journal to assess their normative views on the importance of the academic principal investigator in leading both publicly and industry-funded trials. Results: Several patterns emerged from the Delphi panel surveys. First, panelists viewed involvement of an identified academic principal investigator as most important at the design and planning and the interpretation and dissemination phases of a trial, as compared with the implementation and data collection phase. Second, panelists generally viewed involvement of an identified academic principal investigator as more important in publicly funded than in industry-funded trials. Finally, panelists representing industry stakeholders and United States Food and Drug Administration officials viewed involvement of an identified academic principal investigator as less important, especially for industry-funded trials, than did other groups. Respondents to the normative principal investigator survey generally endorsed the importance of academic principal investigators in leading multicenter randomized controlled trials, both overall (median rating 6 on the 0–6 point scale) and for trial-specific tasks. Both overall and with respect to specific tasks, however, respondents viewed an academic principal investigator’s leadership as more important when considering publicly funded as compared with industry-funded trials. Conclusion: Although members of most stakeholder groups participating in Delphi surveys view involvement of an academic principal investigator with overall responsibility for a multicenter randomized controlled trial as very important, there are notable differences depending on the respondent’s perspective, the specific trial-related task, and the source of funding for the trial. In addition, principal investigators generally view an academic principal investigator as more important to the validity of a publicly funded multicenter randomized controlled trial than to the that of an industry-funded trial. These findings highlight the need to clarify the real-world practices of clinical trial leadership across diverse settings and to assess how those practices align with widely shared norms.

Background: A growing population of older adult cancer survivors faces competing cancer and noncancer health risks. There are limited real-world data on recurrence patterns beyond 5 years post-treatment. Methods: This was a SEER-Medicare retrospective cohort study of patients aged ≥66 with stage I-III breast, colon, or rectal cancer who received definitive surgery and survived ≥5 years from diagnosis without recurrence or second primary malignancy. Late recurrence (5–10 years postdiagnosis) was identified using a validated algorithm to detect treated recurrence in Medicare claims. Demographic and clinical characteristics collected at cancer diagnosis were assessed as predictors of late treated recurrence using restricted mean survival time (RMST) regression. Results: The sample included 12 859 breast, 17 329 colon, and 4427 rectal cancer survivors. The cumulative incidence of late treated recurrence 5-10 years postdiagnosis was 5.0% in breast, 4.4% in colon, and 8.0% in rectal cancer survivors. In all cohorts, stage was associated with shorter RMST. The absolute risk difference between stage I and III was greatest in breast (2% vs 18.1%), followed by rectal (5.2% vs 10.3%) and colon (2.7% vs 6.7%) cancer survivors (P < .001 for all cohorts). Although their effect on RMST was modest (<5%), higher grade, node-positive, and ER-positive disease in breast, left-sided tumors in colon, and radiation in rectal cancer were associated with late treated recurrence. Across all cohorts, the incidence of other-cause mortality (24.1%-34.0%) exceeded cancer-specific mortality (2.9%-6.2%). Conclusions: Late treated recurrence in older long-term survivors is uncommon, but risk remains elevated 5 years postdiagnosis in those with more advanced stage.

Background: Cardiovascular disease (CVD) is the leading cause of noncancer mortality in long-term cancer survivors. Population-level assessment of cancer-related exposures is limited with respect to long-term cardiovascular risk in older survivors who have additional aging-related risks. Methods: This was a Surveillance, Epidemiology, and End Results–Medicare retrospective cohort study of long-term (5-year) survivors of breast, prostate, colon, and rectal cancers who were aged 66 years and older, diagnosed from 2003 to 2012, and received definitive treatment. The primary endpoint was late CVD, defined as myocardial infarction, stroke, congestive heart failure, or cardiomyopathy on an inpatient administrative claim or as Surveillance, Epidemiology, and End Results cause of death occurring 5-15 years postdiagnosis. Restricted mean survival time regression was used to assess predictors of shorter average time without CVD and develop a prediction rule for risk stratification. Survivors were assigned a risk score and stratified into tertiles. Results: Included were 95 100 survivors with a mean age of 74 (6) years at diagnosis. Late CVD occurred in 23.2% of survivors. Older age, comorbidities, and prior CVD were associated with a shorter time without CVD. In contrast, cancer-related factors were not associated, except for stage III breast cancer, and radiation plus androgen deprivation therapy for prostate cancer. Across all cohorts, the high-risk strata had a three- to fourfold higher risk of CVD compared with the low-risk strata. Conclusions: In this cohort of older, long-term cancer survivors, cancer-related exposures were not independently associated with onset of CVD 5-15 years after diagnosis but may still contribute to latent cardiovascular risk. Given the limited impact of cancer-specific factors, cancer-agnostic risk prediction may be adequate to predict individual cardiovascular risk.

Importance: Although incarceration is associated with worse cancer outcomes, little is known about quality of cancer care among people with a history of incarceration.Objective: To describe quality of cancer care among those diagnosed with cancer during or immediately after incarceration.Design, Setting, and Participants: This longitudinal cohort study used Connecticut Tumor Registry data linked to Department of Correction rosters from 2005 through 2016. Patients diagnosed with cancer while incarcerated were matched to those diagnosed within 12 months after incarceration and to those with no history of incarceration. Analysis was performed between March 2024 and January 2025.Exposures: Incarceration status at time of cancer diagnosis.Main Outcomes and Measures: The main outcomes were treatment initiation within 60 days and receipt of recommended pretreatment and therapeutic care, ascertained from medical records and tumor registries. Logistic regression was used to evaluate odds of treatment initiation by 60 days. The proportion of recommended elements of pretreatment and therapeutic care received was calculated, and group differences were assessed using χ² tests.Results: The sample included 690 individuals (233 diagnosed during incarceration, 221 diagnosed within 12 months after incarceration, and 236 with no incarceration history). The mean (SD) age was 49.7 (12.1) years, and 611 (88.6%) were male. Compared with those with no history of incarceration, those diagnosed with cancer during incarceration were less likely to initiate treatment within 60 days (169 of 209 [80.9%] vs 131 of 189 [69.3%]; OR, 0.49 [95% CI, 0.31-0.79]). Initiation of treatment within 60 days was not statistically different for those diagnosed after incarceration vs those with no history of incarceration (144 of 187 [77.0%]; OR, 0.81 [95% CI, 0.49-1.34]). Those diagnosed with local and regional cancers during incarceration received 49.2% of recommended treatment-related care (31 of 63 measures), while those diagnosed after incarceration received 45.3% (34 of 75 measures), compared with 66.7% (36 of 54 measures) among those with no history of incarceration (P < .001). People diagnosed with local and regional cancers during incarceration received 80.6% of recommended pretreatment evaluation (54 of 67 measures), and those diagnosed after received 96.8% (59 of 68 measures), compared with 78.1% (50 of 64 measures) among those with no history of incarceration (P = .01).Conclusions and Relevance: In this cohort study of people with a history of incarceration, incarceration was associated with lower likelihood of receiving timely, high-quality cancer treatment. Along with other published literature, this work suggests that gaps in quality of care may contribute to observed disparities in outcomes.