Validation analysis of a composite real-world mortality endpoint for patients with cancer in the United States

Real-World Clinical Performance of a DNA-Based Comprehensive Genomic Profiling Assay for Detecting Targetable Fusions in Nonsquamous NSCLC

BACKGROUND

Genomic fusions are potent oncogenic drivers across cancer types and many are targetable. We demonstrate the clinical performance of DNA-based comprehensive genomic profiling (CGP) for detecting targetable fusions.

MATERIALS AND METHODS

We analyzed targetable fusion genes in >450 000 tissue specimens profiled using DNA CGP (FoundationOne CDx, FoundationOne). Using a de-identified nationwide (US-based) non-small cell lung cancer (NSCLC) clinico-genomic database, we assessed outcomes in patients with nonsquamous NSCLC (NonSqNSCLC) who received matched therapy based on a fusion identified using DNA CGP. Lastly, we modeled the added value of RNA CGP for fusion detection in NonSqNSCLC.

RESULTS

We observed a broad diversity of fusion partners detected with DNA CGP in conjunction with targetable fusion genes (ALK, BRAF, FGFR2, FGFR3, NTRK1/2/3, RET, and ROS1). In NonSqNSCLC with oncogenic ALK, NTRK, RET, and ROS1 fusions detected by DNA CGP, patients treated with a matched tyrosine kinase inhibitor had better real-world progression-free survival than those receiving alternative treatment regimens and benefit was observed regardless of the results of orthogonal fusion testing. An estimated 1.3% of patients with NonSqNSCLC were predicted to have an oncogenic driver fusion identified by RNA, but not DNA CGP, according to a model that accounts for multiple real-world factors.

CONCLUSION

A well-designed DNA CGP assay is capable of robust fusion detection and these fusion calls are reliable for informing clinical decision-making. While DNA CGP detects most driver fusions, the clinical impact of fusion detection is substantial for individual patients and exhaustive efforts, inclusive of additional RNA-based testing, should be considered when an oncogenic driver is not clearly identified.

Classification and Regression Trees to Predict for Survival for Patients With Hepatocellular Carcinoma Treated With Atezolizumab and Bevacizumab

PURPOSE

Systemic therapy with atezolizumab and bevacizumab can extend life for patients with advanced hepatocellular carcinoma (HCC). However, there is substantial variability in response to therapy and overall survival. Although current prognostic models have been validated in HCC, they primarily consider covariates that may be reflective of the severity of the underlying liver disease of patients with HCC. We developed and internally validated a classification and regression tree (CART) to identify patient characteristics associated with risks of early mortality, at or before 6 months from treatment initiation.

METHODS

This retrospective cohort study used the nationwide Flatiron Health electronic health record-derived deidentified database and included patients with a diagnosis of HCC after January 1, 2020, who received initial systemic therapy with atezolizumab and bevacizumab. CART was developed from available baseline clinical and demographic information to predict mortality within 6 months from treatment initiation. Model characteristics were compared to the albumin-bilirubin (ALBI) model and was further validated against a contemporary validation cohort of patients after a data update.

RESULTS

A total of 293 patients were analyzed. The CART identified seven cohorts of patients from baseline demographic and laboratory characteristics. The model had an area under the receiver operating curve (AUROC) of 0.739 (95% CI, 0.683 to 0.794) for predicting 6-month mortality. This model was internally valid and performed more favorably than the ALBI model, which had an AUROC of 0.608 (95% CI, 0.557 to 0.660). The model applied to the contemporary validation cohort (n = 111) had an AUROC of 0.666 (95% CI, 0.506 to 0.826).

CONCLUSION

Using CART, we identified unique cohorts of patients with HCC treated with atezolizumab and bevacizumab with distinct risks of early mortality. This approach outperformed the ALBI model and used clinical and laboratory characteristics that are readily available to oncologists caring for these patients.

Real-World Treatment Patterns and Clinical Outcomes Among Patients with Metastatic or Unresectable EGFR-Mutated Non-Small Cell Lung Cancer Previously Treated with Osimertinib and Platinum-Based Chemotherapy

INTRODUCTION

For patients with epidermal growth factor receptor-mutated (EGFRm) locally advanced/metastatic non-small cell lung cancer (mNSCLC) whose disease has progressed on or after osimertinib and platinum-based chemotherapy (PBC), no uniformly accepted standard of care exists. Moreover, limited efficacy of standard treatments indicates an unmet medical need, which is being addressed by ongoing clinical investigations, including the HERTHENA-Lung01 (NCT04619004) study of patritumab deruxtecan (HER3‑DXd). However, because limited information is available on real-world clinical outcomes in such patients, early-phase trials of investigational therapies lack sufficient context for comparison. This study describes the real-world clinical characteristics, treatments, and outcomes for patients with EGFRm mNSCLC who initiated a new line of therapy following previous osimertinib and PBC, including a subset matched to the HERTHENA-Lung01 population.

METHODS

This retrospective analysis used a US database derived from deidentified electronic health records. The reference cohort included patients with EGFRm mNSCLC who had initiated a new line of therapy between November 13, 2015 and June 30, 2021, following prior osimertinib and PBC. A subset of patients resembling the HERTHENA-Lung01 population was then extracted from the reference cohort; this matched subset was optimized using propensity score (PS) weighting. Endpoints were real-world overall survival (rwOS) and real-world progression-free survival (rwPFS). Confirmed real-world objective response rate (rwORR; partial/complete response confirmed ≥ 28 days later) was calculated for the response-evaluable subgroups of patients (with ≥ 2 response assessments spaced ≥ 28 days apart).

RESULTS

In the reference cohort (N = 273), multiple treatment regimens were used, and none was predominant. Median rwPFS and rwOS were 3.3 and 8.6 months, respectively; confirmed rwORR (response evaluable, n = 123) was 13.0%. In the matched subset (n = 126), after PS weighting, median rwPFS and rwOS were 4.2 and 9.1 months, respectively; confirmed rwORR (response evaluable, n = 57) was 14.1%.

CONCLUSION

The treatment landscape for this heavily pretreated population of patients with EGFRm mNSCLC is fragmented, with no uniformly accepted standard of care. A high unmet need exists for therapeutic options that provide meaningful improvements in clinical benefit.

Real-World Treatment Patterns and Outcomes of Cemiplimab in Patients with Advanced Cutaneous Squamous Cell Carcinoma Treated in US Oncology Practices

Background

Prior to the Food and Drug Administration approval of cemiplimab in 2018, the median overall survival (OS) for adult patients with advanced CSCC receiving systemic therapy was approximately 8 to 15 months. Limited real-world data are available on cemiplimab for this indication in the US.

Patients and Methods

This retrospective cohort study included US patients with advanced CSCC initiating cemiplimab monotherapy in a real-world database (2018-2021). A clinical trial-like sub-cohort was identified using select criteria. Time to treatment discontinuation (TTD), time to next treatment (TTNT), and OS were estimated using Kaplan-Meier methods. Cox proportional hazard models were used to examine prognostic factors associated with OS in the main cohort.

Results

The main cohort included 622 patients (n = 240 in the trial-like cohort). In the main cohort, the median age was 78 years, 77.8% were male, 21.4% were immunocompromised/immunosuppressed, and 63.8% had metastatic CSCC. Median (95% CI) TTD and TTNT were 8.0 (6.6-9.0) months and 16.4 (13.3-21.0) months, respectively, in the main cohort. Median (95% CI) OS was 24.8 (21.8-29.1) months in the main cohort (not reached in the trial-like cohort). In multivariable analyses, age <60 years (hazard ratio [HR], 0.37), Eastern Cooperative Oncology Group performance status <3-4 (HR range, 0.13-0.57), and primary CSCC location in the head and neck only versus extremities only (HR, 0.59) were associated with better OS. Similar OS was observed between patients who had immunosuppressing/immunocompromising conditions and those without.

Conclusion

These findings confirm the effectiveness of cemiplimab among a heterogenous, real-world advanced CSCC patient population and substantiate the efficacy of cemiplimab observed in clinical trials.

Natural history of SCLC patients treated in third-line and beyond: A retrospective real world study

OBJECTIVES

To describe treatment patterns and estimate outcomes among real-world small cell lung cancer (SCLC) patients in the US who received three or more lines of therapy.

MATERIALS AND METHODS

We conducted a retrospective analysis of adult patients with SCLC who received a front-line platinum-based regimen and two additional lines of therapy (ie., a cohort of at least three lines of therapy). De-identified patients were selected from a United States Flatiron Health oncology database of electronic health records. Treatment patterns were captured by line of therapy. Outcomes evaluated by line of therapy included real-world overall survival (rwOS), real-world progression free survival (rwPFS), real-world response rate (rwRR) and real-world duration of response (rwDOR).

RESULTS

The analysis included 326 3L SCLC patients, of which 103 (32 %) received 4L treatment, and 38 % (39/103) of 4L treated received 5L of therapy. Among the 3L cohort, the average age was 67 years, 49 % were male, and nearly all had a history of smoking (96 %). In the 3L setting, the median rwOS was 5.3 months (95 % Confidence Interval (CI): 4.5, 6.0), median rwPFS was 2.5 months (95 % CI: 2.1, 2.7), rwRR was 19.3 % (95 % CI: 15.2, 24.0) and median DOR was 3.4 months (95 % CI: 2.8, 4.4). No differences were seen in outcomes between the overall cohort and a subgroup of patients treated with front-line platinum-based regimen with an anti-programmed cell death ligand 1 (PD-L1) agent (atezolizumab or durvalumab), in each respective line of therapy.

CONCLUSION

Results from this large, real-world study of US patients with SCLC in the 3L setting and beyond highlight the poor treatment outcomes in advanced SCLC patients with existing therapies and underscore the dire need for new therapies for SCLC patients.

Systemic Anticancer Therapy and Overall Survival in Patients With Very Advanced Solid Tumors

Importance

Two prominent organizations, the American Society of Clinical Oncology and the National Quality Forum (NQF), have developed a cancer quality metric aimed at reducing systemic anticancer therapy administration at the end of life. This metric, NQF 0210 (patients receiving chemotherapy in the last 14 days of life), has been critiqued for focusing only on care for decedents and not including the broader population of patients who may benefit from treatment.

Objective

To evaluate whether the overall population of patients with metastatic cancer receiving care at practices with higher rates of oncologic therapy for very advanced disease experience longer survival.

Design, Setting, and Participants

This nationwide population-based cohort study used Flatiron Health, a deidentified electronic health record database of patients diagnosed with metastatic or advanced disease, to identify adult patients (aged ≥18 years) with 1 of 6 common cancers (breast cancer, colorectal cancer, non-small cell lung cancer [NSCLC], pancreatic cancer, renal cell carcinoma, and urothelial cancer) treated at health care practices from 2015 to 2019. Practices were stratified into quintiles based on retrospectively measured rates of NQF 0210, and overall survival was compared by disease type among all patients treated in each practice quintile from time of metastatic diagnosis using multivariable Cox proportional hazard models with a Bonferroni correction for multiple comparisons. Data were analyzed from July 2021 to July 2023.

Exposure

Practice-level NQF 0210 quintiles.

Main Outcome and Measure

Overall survival.

Results

Of 78 446 patients (mean [SD] age, 67.3 [11.1] years; 52.2% female) across 144 practices, the most common cancer types were NSCLC (34 201 patients [43.6%]) and colorectal cancer (15 804 patients [20.1%]). Practice-level NQF 0210 rates varied from 10.9% (quintile 1) to 32.3% (quintile 5) for NSCLC and 6.8% (quintile 1) to 28.4% (quintile 5) for colorectal cancer. No statistically significant differences in survival were observed between patients treated at the highest and the lowest NQF 0210 quintiles. Compared with patients seen at practices in the lowest NQF 0210 quintiles, the hazard ratio for death among patients seen at the highest quintiles varied from 0.74 (95% CI, 0.55-0.99) for those with renal cell carcinoma to 1.41 (95% CI, 0.98-2.02) for those with urothelial cancer. These differences were not statistically significant after applying the Bonferroni-adjusted critical P = .008.

Conclusions and Relevance

In this cohort study, patients with metastatic or advanced cancer treated at practices with higher NQF 0210 rates did not have improved survival. Future efforts should focus on helping oncologists identify when additional therapy is futile, developing goals of care communication skills, and aligning payment incentives with improved end-of-life care.

Clinical Value of Timely Targeted Therapy for Patients With Advanced Non-Small Cell Lung Cancer With Actionable Driver Oncogenes

BACKGROUND

A recent real-world study observed that 24% of patients with advanced non-small cell lung cancer (aNSCLC) with actionable driver oncogenes (ADOs) initiated nontargeted therapies before biomarker test results became available. This study assessed the clinical impact of the timing of first-line (1L) targeted therapies (TTs) in aNSCLC.

MATERIALS AND METHODS

This retrospective analysis of a nationwide electronic health record-derived deidentified database included patients aged ≥18 years diagnosed with aNSCLC with ADOs (ALK, BRAF, EGFR, RET, MET, ROS-1, and NTRK) from January 1, 2015, to October 18, 2022, by biomarker testing within 90 days after advanced diagnosis and received 1L treatment. Cohorts were defined by treatment patterns ≤42 days after test results: "Upfront TT" received 1L TT ≤42 days; "Switchers" initiated 1L non-TT before or after testing but switched to TT ≤42 days; and "Non-switchers" initiated non-TT before or after testing and did not switch at any time. Adjusted multivariate Cox regression evaluated real-world progression-free survival, real-world time to next treatment or death, and real-world overall survival.

RESULTS

A total of 3540 patients met the study criteria; 78% were treated in a community setting, and 50% underwent next-generation sequencing (NGS). There was no significant difference in outcomes between Switchers and Upfront TT; inferior outcomes were observed in Non-switchers versus Upfront TT.

CONCLUSION

Our findings demonstrated improved outcomes with upfront 1L TT versus non-TT in patients with aNSCLC with ADOs and observed timely switching to TT after biomarker test result had similar outcomes to Upfront TT. Opportunities remain to improve the use of NGS for early ADO identification and determination of 1L TT.

Predictive Impact of Tumor Mutational Burden on Real-World Outcomes of First-Line Immune Checkpoint Inhibition in Metastatic Melanoma

PURPOSE

The choice of threshold and reliability of high tumor mutational burden (TMB) to predict outcomes and guide treatment choice for patients with metastatic melanoma receiving first-line immune checkpoint inhibitor (ICI) therapy in the real world is not well known.

METHODS

Using a deidentified nationwide (US-based) melanoma clinicogenomic database, we identified a real-world cohort of patients with metastatic melanoma (N = 497) who received first-line monotherapy anti-PD-1 (n = 240) or dual anti-PD-1 and anti-CTLA-4 ICI (n = 257) and had a tissue-based comprehensive genomic profiling test TMB score.

RESULTS

TMB-high (TMB-H; ≥10 mutations per megabase [muts/Mb], n = 352, 71%) was independently predictive of superior real-world progression-free survival and overall survival versus TMB-low (<10 mut/Mb, n = 145, 29%) in both mono ICI (hazard ratio [HR], 0.45 [95% CI, 0.32 to 0.63]; P < .001; HR, 0.61 [95% CI, 0.41 to 0.90]; P = .01, respectively) and dual ICI (HR, 0.67 [95% CI, 0.49 to 0.90]; P = .009; HR, 0.61 [95% CI, 0.42 to 0.88]; P = .007, respectively) patients. Dual ICI offered no significant advantage in BRAFwt patients and unexpectedly demonstrated greatest benefit in the TMB 10-19 mut/Mb group, identifying a TMB-very high (≥20 mut/Mb, n = 247, 50%) BRAFmut patient subgroup for whom mono ICI may be preferable.

CONCLUSION

TMB-H predicts superior outcomes on ICI while coassessment of BRAF status and TMB may inform first-line regimen choice.

Comparative effectiveness of lenalidomide/dexamethasone-based triplet regimens for treatment of relapsed and/or refractory multiple myeloma in the United States: An analysis of real-world electronic health records data

BACKGROUND

This retrospective longitudinal study compared the effectiveness of dexamethasone+lenalidomide (Rd)-based triplet regimens containing proteasome inhibitors (PIs) ixazomib (IRd), carfilzomib (KRd), and bortezomib (VRd) or monoclonal antibodies (MABs) elotuzumab (ERd) and daratumumab (DRd) in patients with relapsed/refractory multiple myeloma (RRMM)-including those with high cytogenetic risk-primarily treated at community oncology clinics in the United States.

METHODS

Electronic health records of adult RRMM patients in a deidentified real-world database (01/01/2014-09/30/2020) who initiated IRd, KRd, VRd, ERd, or DRd in the second or later line of therapy (LOT) were analyzed. The index date was the date of initiation of each LOT and baseline was the 6-month pre-index period. Duration of therapy (DOT), time to next therapy (TTNT), progression-free survival (PFS), and overall survival (OS) were compared across regimens with multivariable Cox proportional hazards models.

RESULTS

Of the 1,185 patients contributing 1,332 LOTs, 985 had standard cytogenetic risk (median age, 71 years) and 180 had high risk (median age, 69 years). Compared with other regimens, DRd was associated with longer DOT overall (adjusted hazard ratio [95 % confidence interval]: 1.84 [1.42, 2.38] vs. KRd, 1.65 [1.20, 2.28] vs. ERd, 1.58 [1.23, 2.04] vs. IRd, and 1.54 [1.18, 2.00] vs. VRd), and longer TTNT and PFS. KRd was associated with shorter OS compared with DRd (1.45 [1.01, 2.08]) and VRd (1.32 [1.01, 1.73]). High-risk patients had similar outcomes with all triplet regimens.

CONCLUSION

Although DRd improved clinical outcomes overall, Rd-based triplet regimens containing a PI or MAB are similarly effective in high-risk RRMM.

5-Year Real-World Outcomes With Frontline Pembrolizumab Monotherapy in PD-L1 Expression ≥ 50% Advanced NSCLC

BACKGROUND

In clinical trials, frontline pembrolizumab for advanced NSCLC has demonstrated durable, clinically meaningful, long-term survival benefits over chemotherapy. Our objective was to evaluate 5-year survival rates outside the idealized setting of clinical trials for advanced/metastatic NSCLC treated with frontline pembrolizumab monotherapy.

METHODS

Using a nationwide, electronic health record-derived, deidentified database in the United States, we studied adult patients with advanced/metastatic NSCLC (unresectable stage IIIB/IIIC, or stage IV), with PD-L1 expression ≥ 50%, no documented EGFR, ALK, or ROS1 genomic alteration, and ECOG performance status of 0-1 initiating frontline pembrolizumab monotherapy from November 1, 2016, through March 31, 2020, excluding those in clinical trials. Kaplan-Meier was used to determine overall survival (OS). Data cutoff was May 31, 2023.

RESULTS

A total of 804 patients were eligible for the study, including 404 women (50%); median age was 72 years (range, 38-85 years), with 310 patients (39%) ≥ 75 years old. Median follow-up time from pembrolizumab initiation to data cutoff was 60.5 months (range, 38.0-78.7). At data cutoff, 549 patients (68%) had died. Median OS was 19.2 months (95% CI, 16.6-21.4), and survival rate at 5 years was 25.1% (95% CI, 21.7-28.7). Overall, 266 patients (33%) received 1 or more subsequent regimens, most commonly an anti-PD-(L)1 agent (as monotherapy or combination therapy) or platinum-based chemotherapy.

CONCLUSIONS

With 5-year follow-up in a real-world population, frontline pembrolizumab monotherapy continues to demonstrate long-term effectiveness, with survival outcomes consistent with those of pivotal clinical trials, for treating patients with advanced NSCLC with PD-L1 expression of ≥ 50% and no EGFR, ALK, or ROS1 genomic alteration.

Clinical outcomes of adjuvant nivolumab in resected stage III melanoma: comparison of CheckMate 238 trial and real-world data

OBJECTIVES

Nivolumab is approved as adjuvant therapy for resected stage III/IV melanoma based on the phase 3 CheckMate 238 trial. This analysis compared outcomes from CheckMate 238 with those from the real-world Flatiron Health electronic health record-derived de-identified database in patients with resected stage III melanoma (per AJCC-8) treated with adjuvant nivolumab.

MATERIALS

Outcomes included baseline characteristics, overall survival (OS) in the CheckMate 238 cohort (randomization until death or last known alive), and real-world overall survival (rwOS) in the Flatiron Health cohort (nivolumab initiation until death or data cutoff). rwOS was compared with OS using unadjusted and adjusted Cox proportional hazards models. Inverse probability of treatment weighting (IPTW) was combined with the adjusted model to reduce baseline discrepancies.

RESULTS

The CheckMate 238 and real-world cohorts included 369 and 452 patients, respectively (median age, 56.0 and 63.0 years; median follow-up, 61.4 vs. 25.5 months). rwOS was not different from OS in the unadjusted (hazard ratio [HR] 1.27; 95% CI 0.92-1.74), adjusted (HR 1.01; 95% CI 0.67-1.54), and adjusted IPTW (HR 1.07; 95% CI 0.70-1.63) analyses. In the adjusted analysis, 2-year OS and rwOS rates were 84%. Median OS and rwOS were not reached. After IPTW, OS and rwOS were not different (HR 1.07; 95% CI 0.70-1.64).

CONCLUSIONS

In this comparative analysis, OS in the CheckMate 238 trial was similar to rwOS in the Flatiron Health database after adjustments in patients with resected stage III melanoma (per AJCC-8) treated with adjuvant nivolumab, validating the trial results.

Sex Differences in Odds of Brain Metastasis and Outcomes by Brain Metastasis Status after Advanced Melanoma Diagnosis

Sex differences in cancer are well-established. However, less is known about sex differences in diagnosis of brain metastasis and outcomes among patients with advanced melanoma. Using a United States nationwide electronic health record-derived de-identified database, we evaluated patients diagnosed with advanced melanoma from 1 January 2011-30 July 2022 who received an oncologist-defined rule-based first line of therapy (n = 7969, 33% female according to EHR, 35% w/documentation of brain metastases). The odds of documented brain metastasis diagnosis were calculated using multivariable logistic regression adjusted for age, practice type, diagnosis period (pre/post-2017), ECOG performance status, anatomic site of melanoma, group stage, documentation of non-brain metastases prior to first-line of treatment, and BRAF positive status. Real-world overall survival (rwOS) and progression-free survival (rwPFS) starting from first-line initiation were assessed by sex, accounting for brain metastasis diagnosis as a time-varying covariate using the Cox proportional hazards model, with the same adjustments as the logistic model, excluding group stage, while also adjusting for race, socioeconomic status, and insurance status. Adjusted analysis revealed males with advanced melanoma were 22% more likely to receive a brain metastasis diagnosis compared to females (adjusted odds ratio [aOR]: 1.22, 95% confidence interval [CI]: 1.09, 1.36). Males with brain metastases had worse rwOS (aHR: 1.15, 95% CI: 1.04, 1.28) but not worse rwPFS (adjusted hazard ratio [aHR]: 1.04, 95% CI: 0.95, 1.14) following first-line treatment initiation. Among patients with advanced melanoma who were not diagnosed with brain metastases, survival was not different by sex (rwOS aHR: 1.06 [95% CI: 0.97, 1.16], rwPFS aHR: 1.02 [95% CI: 0.94, 1.1]). This study showed that males had greater odds of brain metastasis and, among those with brain metastasis, poorer rwOS compared to females, while there were no sex differences in clinical outcomes for those with advanced melanoma without brain metastasis.

A matching-based machine learning approach to estimating optimal dynamic treatment regimes with time-to-event outcomes

Observational data (e.g. electronic health records) has become increasingly important in evidence-based research on dynamic treatment regimes, which tailor treatments over time to patients based on their characteristics and evolving clinical history. It is of great interest for clinicians and statisticians to identify an optimal dynamic treatment regime that can produce the best expected clinical outcome for each individual and thus maximize the treatment benefit over the population. Observational data impose various challenges for using statistical tools to estimate optimal dynamic treatment regimes. Notably, the task becomes more sophisticated when the clinical outcome of primary interest is time-to-event. Here, we propose a matching-based machine learning method to identify the optimal dynamic treatment regime with time-to-event outcomes subject to right-censoring using electronic health record data. In contrast to the established inverse probability weighting-based dynamic treatment regime methods, our proposed approach provides better protection against model misspecification and extreme weights in the context of treatment sequences, effectively addressing a prevalent challenge in the longitudinal analysis of electronic health record data. In simulations, the proposed method demonstrates robust performance across a range of scenarios. In addition, we illustrate the method with an application to estimate optimal dynamic treatment regimes for patients with advanced non-small cell lung cancer using a real-world, nationwide electronic health record database from Flatiron Health.

A Quantitative Bias Analysis Approach to Informative Presence Bias in Electronic Health Records

Accurate outcome and exposure ascertainment in electronic health record (EHR) data, referred to as EHR phenotyping, relies on the completeness and accuracy of EHR data for each individual. However, some individuals, such as those with a greater comorbidity burden, visit the health care system more frequently and thus have more complete data, compared with others. Ignoring such dependence of exposure and outcome misclassification on visit frequency can bias estimates of associations in EHR analysis. We developed a framework for describing the structure of outcome and exposure misclassification due to informative visit processes in EHR data and assessed the utility of a quantitative bias analysis approach to adjusting for bias induced by informative visit patterns. Using simulations, we found that this method produced unbiased estimates across all informative visit structures, if the phenotype sensitivity and specificity were correctly specified. We applied this method in an example where the association between diabetes and progression-free survival in metastatic breast cancer patients may be subject to informative presence bias. The quantitative bias analysis approach allowed us to evaluate robustness of results to informative presence bias and indicated that findings were unlikely to change across a range of plausible values for phenotype sensitivity and specificity. Researchers using EHR data should carefully consider the informative visit structure reflected in their data and use appropriate approaches such as the quantitative bias analysis approach described here to evaluate robustness of study findings.

Real-world outcomes on platinum-containing chemotherapy for EGFR-mutated advanced nonsquamous NSCLC with prior exposure to EGFR tyrosine kinase inhibitors

Background

Front-line therapy with an EGFR tyrosine kinase inhibitor (TKI) is the standard of care for treating patients with advanced nonsquamous NSCLC with the common sensitizing EGFR exon 19 deletion and exon 21 L858R point mutations. However, EGFR TKI resistance inevitably develops. The optimal subsequent therapy remains to be identified, although platinum-containing chemotherapy regimens are often administered. Our objectives were to describe baseline characteristics, survival, and subsequent treatment patterns for patients with advanced nonsquamous NSCLC with EGFR exon 19 deletion or L858R mutation who received a platinum-based combination regimen after front-line EGFR TKI therapy.

Methods

This retrospective study used a nationwide electronic health record-derived deidentified database to select adult patients with advanced nonsquamous NSCLC, evidence of EGFR exon 19 deletion or L858R mutation, and ECOG performance status of 0-2 who initiated platinum-containing chemotherapy, with or without concomitant immunotherapy, from 1-January-2011 to 30-June-2020 following receipt of any EGFR TKI as first-line therapy or, alternatively, a first- or second-generation EGFR TKI (erlotinib, afatinib, gefitinib, dacomitinib) as first-line therapy followed by the third-generation EGFR TKI osimertinib as second-line therapy. Data cut-off was 30-June-2022. The Kaplan-Meier method was used to estimate overall survival (OS) after initiation of pemetrexed-platinum (n=119) or any platinum-based combination regimen (platinum cohort; n=311).

Results

The two cohorts included two-thirds women (65%-66%) and 57%-58% nonsmokers; median ages were 66 and 65 years in pemetrexed-platinum and platinum cohorts, respectively. Median OS was 10.3 months (95% CI, 8.1-13.9) from pemetrexed-platinum initiation and 12.4 months (95% CI, 10.2-15.2) from platinum initiation; 12-month survival rates were 48% and 51%, respectively; 260 patients (84%) had died by the end of the study.

Conclusion

The suboptimal survival outcomes recorded in this study demonstrate the unmet need to identify more effective subsequent treatment regimens for patients with EGFR-mutated advanced nonsquamous NSCLC after EGFR TKI resistance develops.

Utility of real-world evidence in biosimilar development

Biosimilar development refers to the process of creating a biologic drug that is similar to an existing approved biologic drug, also known as a reference drug. Due to the complex nature of biologics drugs and the inherent variability in their manufacturing process biosimilars are not identical but highly similar to the reference drug in terms of quality, safety, and efficacy. Efficacy and safety trials for biosimilars involve large numbers of patients to confirm comparable clinical performance of the biosimilar and the reference product in appropriately sensitive clinical indications and for appropriate sensitive endpoints. The objective of a biosimilar clinical data is to address slight differences observed at previous steps and to confirm comparable clinical performance of the biosimilar and the reference product. In recent years with advances in big data computing, there has been increasing interest to incorporate the totality of information from different data sources (e.g. Real World data and published literature) in design and conduct of clinical trial to support regulatory objectives. The biosimilar development is an ideal framework for utilization of Real-World Evidence in design of trials as potentially large amount of data are available for the reference dug. Hence there may be an opportunity to use RWD in establishing, improving or validating equivalence margins (EQM) for biosimilar designs, specifically in the case there is no historical published data in the intended sensitive population. In this article, we propose a variation of matching method that seems promising to identify the matched set from a real-world data for which the effect size of targeted endpoint would be comparable to historical data. We believe this is a reasonable approach because in design stage, we can view covariates and secondary endpoints as data feature that can be used in a matching method. This approach was illustrated through a case study which indicated the estimate of the primary endpoint is within 1% of published results and thus RWD may be used to justify or estimate the equivalence margin. To ensure consistent results we recommend using this approach in different indications and endpoint scenarios. Thus utilization of RWD/RWE can provide an important opportunity to increase access to biologic therapies, reducing cost by repurposing existing data.

RAS/RAF Comutation and ERBB2 Copy Number Modulates HER2 Heterogeneity and Responsiveness to HER2-directed Therapy in Colorectal Cancer

PURPOSE

ERBB2-amplified colorectal cancer is a distinct molecular subtype with expanding treatments. Implications of concurrent oncogenic RAS/RAF alterations are not known.

EXPERIMENTAL DESIGN

Dana-Farber and Foundation Medicine Inc. Colorectal cancer cohorts with genomic profiling were used to identify ERBB2-amplified cases [Dana-Farber, n = 47/2,729 (1.7%); FMI, n = 1857/49,839 (3.7%)]. Outcomes of patients receiving HER2-directed therapies are reported (Dana-Farber, n = 9; Flatiron Health-Foundation Medicine clinicogenomic database, FH-FMI CGDB, n = 38). Multisite HER2 IHC and genomic profiling were performed to understand HER2 intratumoral and interlesional heterogeneity. The impact of concurrent RAS comutations on the effectiveness of HER2-directed therapies were studied in isogenic colorectal cancer cell lines and xenografts.

RESULTS

ERBB2 amplifications are enriched in left-sided colorectal cancer. Twenty percent of ERBB2-amplified colorectal cancers have co-occurring oncogenic RAS/RAF alterations. While RAS/RAF WT colorectal cancers typically have clonal ERBB2 amplification, colorectal cancers with co-occurring RAS/RAF alterations have lower level ERRB2 amplification, higher intratumoral heterogeneity, and interlesional ERBB2 discordance. These distinct genomic patterns lead to differential responsiveness and patterns of resistance to HER2-directed therapy. ERBB2-amplified colorectal cancer with RAS/RAF alterations are resistant to trastuzumab-based combinations, such as trastuzumab/tucatinib, but retain sensitivity to trastuzumab deruxtecan in in vitro and murine models. Trastuzumab deruxtecan shows clinical efficacy in cases with high-level ERBB2-amplified RAS/RAF coaltered colorectal cancer.

CONCLUSIONS

Co-occurring RAS/RAF alterations define a unique subtype of ERBB2-amplified colorectal cancer that has increased intratumoral heterogeneity, interlesional discordance, and resistance to trastuzumab-based combinations. Further examination of trastuzumab deruxtecan in this previously understudied cohort of ERBB2-amplified colorectal cancer is warranted.

Real-World Outcomes of Trastuzumab Deruxtecan in Patients With HER2+ Metastatic Breast Cancer: The DE-REAL Study

BACKGROUND

Trastuzumab deruxtecan (T-DXd) demonstrated unprecedented efficacy in patients with pretreated HER2+ metastatic breast cancer (mBC). However, few data are available about its efficacy in routine clinical practice. In this multicenter retrospective study, we examined effectiveness and safety of T-DXd in a real-world population.

METHODS

Clinico-pathological information about patients with HER2+ mBC who received T-DXd were collected from 12 Italian hospitals. HER2 status was determined locally. Patients who received at least one administration of T-DXd, as any therapy line for advanced disease were included in the analysis. The primary endpoint was real-word PFS (rwPFS).

RESULTS

One hundred and forty-three patients were included. Median age was 66 (range: 37-90), and 4 men were included. Hormone receptor (HR) status was positive in 108 (75%) patients and negative in 35(25%). T-DXd was administered as first, second, third, or subsequent lines in 4 (3%), 16 (11%), 42 (29%), and 81 (57%) patients, respectively. Among 123 patients with measurable disease, the ORR was 68%, and the DCR was 93% (9 CRs, 74 PRs, and 30 SD). Nine (7%) patients had a primary resistance to T-DXd. With a median follow-up of 12 months, the median rwPFS was 16 months. RwPFS was 84%, 59%, and 39% at 6, 12, and 18 months, respectively. A favorable trend in rwPFS was reported in patients receiving T-DXd as I/II line versus further lines (17 vs. 15 months; P = .098). Any-grade toxicity was registered in 84 patients (59%). Most common adverse events (AEs) reported were nausea (33%), neutropenia (21%), and asthenia (21%). Liver toxicity and diarrhea were uncommon (5% and 1%). Severe toxicities was registered in 18% of patients, and the most frequent were neutropenia, nausea/vomiting, and ILD observed in 15, 2, and 3 patients. AEs led to dose reduction in 37 patients (26%). Dose reduction and AEs do not affect patients' response and survival outcomes.

CONCLUSIONS

Efficacy and safety of T-DXd were confirmed in an unselected real-world population of HER2+ mBC. These results are consistent with the results of known findings, and no new safety concerns were reported.

Outcomes in patients treated with frontline immune checkpoint inhibition (ICI) for advanced NSCLC with KRAS mutations and STK11/KEAP1 comutations across PD-L1 levels

INTRODUCTION

In patients with advanced NSCLC (aNSCLC), the impact of KRAS mutations (m) and comutations with STK11 and KEAP1 on outcomes across different PD-L1 levels remains incompletely understood. We aimed to investigate the frequency of KRAS mutations and comutations across PD-L1 levels, and the association between these mutations and survival, stratified by PD-L1 expression.

METHODS

We conducted a nationwide cohort study of patients diagnosed with aNSCLC between 2016 and 2021 treated with frontline (chemo)immunotherapy, who underwent molecular genotyping including KRAS, STK11, and KEAP1. Real-world overall survival (OS) and progression-free survival (rwPFS) were estimated using Kaplan-Meier methodology. Cox multivariable regressions were used to evaluate the association between KRASm and survival across different PD-L1 strata, and to assess whether the association between KRASm and survival differed by PD-L1 level. Finally, within subgroups defined by PD-L1 expression, we used interaction terms to assess whether co-mutations with STK11 and KEAP1 moderated the association between KRAS mutation and survival.

RESULTS

Of our 2593-patient cohort, 982 (37.9 %) were KRASm and 1611 (62.1 %) KRASwt. KRASm were enriched in the PD-L1 ≥50 % cohort (334/743, 45 %), but within patients with KRASm, co-mutations with STK11 and KEAP1 were enriched in the PD-L1 0 % cohort. KRASm was associated with significantly worse OS in the PD-L1 0 % cohort compared to the PD-L1 ≥50 % cohort (P for interaction = 0.008). On adjusted analyses stratified by PD-L1, KRASm was associated with worse survival only in the PD-L1 0 % group (OS HR 1.46, p = 0.001). KEAP1 and STK11 comutations were most strongly associated with worse OS in the PD-L1 0 % subgroup; patients with triple KRASm/KEAPm/STK11m PD-L1 0 % NSCLC experienced the worst outcomes.

CONCLUSIONS

KRASm are associated with worse overall survival in PD-L1 negative NSCLC; however, this association is largely driven by comutations with STK11 and KEAP1, which are enriched in PD-L1 negative tumors.

Broadening Eligibility Criteria and Diversity among Patients for Cancer Clinical Trials

BACKGROUND

Certain populations have been historically underrepresented in clinical trials. Broadening eligibility criteria is one approach to inclusive clinical research and achieving enrollment goals. How broadened trial eligibility criteria affect the diversity of eligible participants is unknown.

METHODS

Using a nationwide electronic health record-derived deidentified database, we identified a retrospective cohort of patients diagnosed with 22 cancer types between April 1, 2013 and December 31, 2022 who received systemic therapy (N=235,234) for cancer. We evaluated strict versus broadened eligibility criteria using performance status and liver, kidney, and hematologic function around first line of therapy. We performed logistic regression to estimate odds ratios for exclusion by strict criteria and their association with measures of patient diversity, including sex, age, race or ethnicity, and area-level socioeconomic status (SES); estimated the impact of broadening criteria on the number and distribution of eligible patients; and performed Cox regression to estimate hazard ratios for real-world overall survival (rwOS) comparing patients meeting strict versus broadened criteria.

RESULTS

When applying common strict cutoffs for eligibility criteria to patients with complete data and weighting each cancer type equally, 48% of patients were eligible for clinical trials. Female (odds ratio, 1.30; 95% confidence interval [CI], 1.25 to 1.35), older (age 75+ vs. 18 to 49 years old: odds ratio, 3.04; 95% CI, 2.85 to 3.24), Latinx (odds ratio, 1.46; 95% CI, 1.39 to 1.54), non-Latinx Black (odds ratio, 1.11; 95% CI, 1.06 to 1.16), and lower-SES patients were more likely to be excluded using strict eligibility criteria. Broadening criteria increased the number of eligible patients by 78%, with the strongest impact for older, female, non-Latinx Black, and lower-SES patients. Patients who met only broadened criteria had worse rwOS versus those with strict criteria (hazard ratio, 1.31; 95% CI, 1.27 to 1.34).

CONCLUSIONS

Data-driven evaluation of clinical trial eligibility criteria may optimize the eligibility of certain historically underrepresented groups and promote access to more inclusive trials. (Sponsored by Flatiron Health.).

Once-weekly versus twice-weekly bortezomib in newly diagnosed multiple myeloma: a real-world analysis

Induction regimens for multiple myeloma (MM) commonly include bortezomib, which has typically been administered twice weekly despite studies demonstrating comparable efficacy and less peripheral neuropathy (PN) with once-weekly bortezomib. We aimed to analyze the real-world prevalence and efficacy of once-weekly versus twice-weekly bortezomib regimens in newly diagnosed MM. We analyzed 2497 US patients aged 18-70 years treated with commercial first-line bortezomib using nationwide Flatiron Health electronic health record-derived data, including 910 (36.4%) patients who received twice-weekly and 1522 (63.2%) who received once-weekly bortezomib. Once-weekly bortezomib use increased over time, from 57.7% in 2017 to 73.1% in 2022. Multivariate analysis identified worsened performance status and more recent year of diagnosis with higher odds of receiving once-weekly bortezomib. Real-world progression-free survival (median 37.2 months with once-weekly versus 39.6 months with twice-weekly, p = 0.906) and overall survival (medians not reached in either cohort, p = 0.800) were comparable. PN rates were higher in patients receiving twice-weekly bortezomib (34.7% versus 18.5%, p < 0.001). In conclusion, once-weekly bortezomib is clearly associated with similar efficacy and fewer toxicities compared to twice-weekly bortezomib. Our findings support once-weekly bortezomib as a standard-of-care regimen for newly diagnosed patients with MM.

Systematic analysis of off-label and off-guideline cancer therapy usage in a real-world cohort of 165,912 US patients

Patients with cancer may be given treatments that are not officially approved (off-label) or recommended by guidelines (off-guideline). Here we present a data science framework to systematically characterize off-label and off-guideline usages using real-world data from de-identified electronic health records (EHR). We analyze treatment patterns in 165,912 US patients with 14 common cancer types. We find that 18.6% and 4.4% of patients have received at least one line of off-label and off-guideline cancer drugs, respectively. Patients with worse performance status, in later lines, or treated at academic hospitals are significantly more likely to receive off-label and off-guideline drugs. To quantify how predictable off-guideline usage is, we developed machine learning models to predict which drug a patient is likely to receive based on their clinical characteristics and previous treatments. Finally, we demonstrate that our systematic analyses generate hypotheses about patients' response to treatments.

Association of Timely Comprehensive Genomic Profiling With Precision Oncology Treatment Use and Patient Outcomes in Advanced Non-Small-Cell Lung Cancer

PURPOSE

Timely biomarker testing remains out of reach for many patients with advanced non-small-cell lung cancer (aNSCLC). Here, we studied the quality-of-care implications of closing the gap in timely receipt of comprehensive genomic profiling (CGP) to inform first-line (1L) decisions.

METHODS

Using a real-world clinicogenomic database, we studied testing and 1L treatment patterns in aNSCLC after the approval of pembrolizumab in combination with pemetrexed and carboplatin (May 10, 2017). To estimate the association of timely CGP results with therapy selection and patient outcomes, we identified patients with no previous genomic testing beyond PD-L1 immunohistochemistry and dichotomized patients by whether CGP results were available before or after 1L therapy initiation.

RESULTS

In total, 2,694 patients were included in the 1L therapy decision impact assessment. Timely CGP increased matched targeted therapy use by 14 percentage points (17% with CGP v 2.8% without) and precision immune checkpoint inhibitor (ICPI) use by 14 percentage points (18% with CGP v 3.9% without). Receipt of timely CGP resulted in an estimated 31 percentage point decrease in ICPI use among ALK/EGFR/RET/ROS1-positive patients at an expected per-patient reduction in ineffective ICPI therapy cost of $13,659.37 with timely CGP to inform 1L treatment selection. Patient benefit of CGP extended to real-world time to therapy discontinuation (median time to therapy discontinuation: 3.9 v 10 months [hazard ratio, HR, 0.54 [95% CI, 0.42 to 0.70]; P = 1.9E-06; adjusted hazard ratio [aHR], 0.50 [95% CI, 0.38 to 0.67]; P = 2.0E-06) in 1L driver-positive patients. This effect was not significant for real-world overall survival (median overall survival: 32 v 29 months [HR, 1.2 [95% CI, 0.84 to 1.67]; P = .33; aHR, 1.4 [95% CI, 0.92 to 1.99]; P = .12).

CONCLUSION

Timely CGP is associated with the quality of patient care as measured by 1L matched targeted therapy use, time to therapy discontinuation, and avoidance of ineffective, costly ICPIs.

Real-World Overall Survival and Treatment Patterns by PTEN Status in Metastatic Castration-Resistant Prostate Cancer

PURPOSE

It is estimated that the PTEN tumor suppressor gene is functionally lost in 40%-50% of patients with metastatic castration-resistant prostate cancer (mCRPC). There is limited information on the prognostic significance of PTEN status identified with genomic testing. This real-world cohort study assessed PTEN as a genetic biomarker using data from US-based oncology practices.

METHODS

This retrospective real-world cohort study used a deidentified US-based metastatic prostate cancer clinicogenomic database linked to longitudinal clinical data derived from electronic health records. Patients were aged 18 years and older and diagnosed with mCRPC between January 1, 2018, and June 30, 2021. Comprehensive genomic profiling (CGP) of tumor specimens was performed using next-generation sequencing. First-line (1L) and second-line (2L) treatment patterns were assessed and stratified by PTEN status. Kaplan-Meier methods and a multivariable Cox model were used to compare the real-world overall survival by PTEN status among patients who received 1L novel hormone therapy or taxanes.

RESULTS

In patients with mCRPC who underwent CGP, PTEN loss of function (LOF) was associated with decreased survival compared with intact PTEN (hazard ratio, 1.61 [95% CI, 1.07 to 2.42]; P = .024). The results were not influenced by 1L treatment type. 1L treatment patterns were similar between intact PTEN and PTEN LOF subgroups, with abiraterone and enzalutamide being the two most common treatments in both groups. Patients with PTEN LOF were less likely to receive 2L treatments than patients with intact PTEN.

CONCLUSION

PTEN LOF, identified with genomic testing, was associated with decreased survival and negative prognoses in patients with mCRPC.

Raising the Bar for Real-World Data in Oncology: Approaches to Quality Across Multiple Dimensions

PURPOSE

Electronic health record (EHR)-based real-world data (RWD) are integral to oncology research, and understanding fitness for use is critical for data users. Complexity of data sources and curation methods necessitate transparency into how quality is approached. We describe the application of data quality dimensions in curating EHR-derived oncology RWD.

METHODS

A targeted review was conducted to summarize data quality dimensions in frameworks published by the European Medicines Agency, The National Institute for Healthcare and Excellence, US Food and Drug Administration, Duke-Margolis Center for Health Policy, and Patient-Centered Outcomes Research Institute. We then characterized quality processes applied to curation of Flatiron Health RWD, which originate from EHRs of a nationwide network of academic and community cancer clinics, across the summarized quality dimensions.

RESULTS

The primary quality dimensions across frameworks were relevance (including subdimensions of availability, sufficiency, and representativeness) and reliability (including subdimensions of accuracy, completeness, provenance, and timeliness). Flatiron Health RWD quality processes were aligned to each dimension. Relevancy to broad or specific use cases is optimized through data set size and variable breadth and depth. Accuracy is addressed using validation approaches, such as comparison with external or internal reference standards or indirect benchmarking, and verification checks for conformance, consistency, and plausibility, selected on the basis of feasibility and criticality of the variable to the intended use case. Completeness is assessed against expected source documentation; provenance by recording data transformation, management procedures, and auditable metadata; and timeliness by setting refresh frequency to minimize data lags.

CONCLUSION

Development of high-quality, scaled, EHR-based RWD requires integration of systematic processes across the data lifecycle. Approaches to quality are optimized through knowledge of data sources, curation processes, and use case needs. By addressing quality dimensions from published frameworks, Flatiron Health RWD enable transparency in determining fitness for real-world evidence generation.

Sociodemographic associations with uptake of novel therapies for acute myeloid leukemia

Inequitable uptake of novel therapies (NT) in non-cancer settings are known for patients with lower socioeconomic status (SES), People of Color (POC), and older adults. NT uptake equity in acute myeloid leukemia (AML) is not well known. We performed a retrospective cohort study (1/2014-8/2022) of the United States nationwide Flatiron HealthTM electronic health record-derived, de-identified database. We estimated sociodemographic associations with AML NT receipt using incidence rate ratios (IRR). Odds ratios (OR) assessed differences in venetoclax (the most common NT) receipt at community sites and between site characteristics and NT adoption. Of 8081 patients (139 sites), 3102 (38%) received a NT. NT use increased annually (IRR 1.14, 95% confidence interval [1.07, 1.22]). NT receipt was similar between Non-Hispanic-Whites and POC (IRR 1.03, [0.91, 1.17]) and as age increased (IRR 1.02 [0.97, 1.07]). At community sites, Non-Hispanic-Whites were less likely to receive venetoclax (OR 0.77 [0.66, 0.91]); older age (OR 1.05 [1.04, 1.05]) and higher area-level SES were associated with venetoclax receipt (OR 1.23 [1.05, 1.43]). Early NT adopting sites had more prescribing physicians (OR 1.25 [1.13, 1.43]) and higher SES strata patients (OR 2.81 [1.08, 7.66]). Inequities in AML NT uptake were seen by SES; for venetoclax, differential uptake reflects its label indication for older adults and those with comorbidities.

Real-World Progression-Free Survival as an Endpoint in Lung Cancer: Replicating Atezolizumab and Docetaxel Arms of the OAK Trial Using Real-World Data

Evaluating cancer treatments in real-world data (RWD) requires informative endpoints. This study replicated the atezolizumab and docetaxel arms of the OAK trial using RWD and compared progression-free survival (PFS) outcomes derived from abstracted physician's notes in RWD (rwPFS) against PFS outcomes derived from the clinical trial PFS (ctPFS). Atezolizumab and docetaxel arms of the phase III OAK randomized controlled trial (RCT; NCT02008227) were replicated in a US nationwide real-world database using selected OAK inclusion/exclusion criteria and propensity score-based adjustment for baseline prognostic variables. Concordance of outcomes was assessed using Kaplan-Meier medians and hazard ratios (HRs). The RWD cohorts comprised 133 patients on atezolizumab and 479 patients on docetaxel. After adjustment, prognostic variables were balanced between RCT arms and corresponding RWD cohorts. The rwPFS and ctPFS outcomes showed better concordance for docetaxel (2.99 vs. 3.52 months; HR: 0.99, 95% confidence interval (CI): 0.85-1.15) than for atezolizumab (3.71 vs. 2.76 months; HR: 0.8, 95% CI: 0.61-1.02). Excluding events labeled "pseudo-progression" from both RWD and RCT improved concordance for atezolizumab (4.24 vs. 4.14 months; HR: 0.95, 95% CI: 0.70-1.25). These findings were robust across sensitivity analyses. Replicating RCTs using RWD and comparing outcomes can help characterize RWD endpoints. Similarity of results between rwPFS and ctPFS at the cohort level may depend on drug category, highlighting the need for further studies to verify and understand when the corresponding outcomes can be compared, including within the same patient.

Creation of a data commons for substance misuse related health research through privacy-preserving patient record linkage between hospitals and state agencies

Objectives

Substance misuse is a complex and heterogeneous set of conditions associated with high mortality and regional/demographic variations. Existing data systems are siloed and have been ineffective in curtailing the substance misuse epidemic. Therefore, we aimed to build a novel informatics platform, the Substance Misuse Data Commons (SMDC), by integrating multiple data modalities to provide a unified record of information crucial to improving outcomes in substance misuse patients.

Materials and Methods

The SMDC was created by linking electronic health record (EHR) data from adult cases of substance (alcohol, opioid, nonopioid drug) misuse at the University of Wisconsin hospitals to socioeconomic and state agency data. To ensure private and secure data exchange, Privacy-Preserving Record Linkage (PPRL) and Honest Broker services were utilized. The overlap in mortality reporting among the EHR, state Vital Statistics, and a commercial national data source was assessed.

Results

The SMDC included data from 36 522 patients experiencing 62 594 healthcare encounters. Over half of patients were linked to the statewide ambulance database and prescription drug monitoring program. Chronic diseases accounted for most underlying causes of death, while drug-related overdoses constituted 8%. Our analysis of mortality revealed a 49.1% overlap across the 3 data sources. Nonoverlapping deaths were associated with poor socioeconomic indicators.

Discussion

Through PPRL, the SMDC enabled the longitudinal integration of multimodal data. Combining death data from local, state, and national sources enhanced mortality tracking and exposed disparities.

Conclusion

The SMDC provides a comprehensive resource for clinical providers and policymakers to inform interventions targeting substance misuse-related hospitalizations, overdoses, and death.

Palbociclib Combined with an Aromatase Inhibitor in Patients with Breast Cancer with Lung or Liver Metastases in US Clinical Practice

A cyclin-dependent kinase 4/6 inhibitor combined with endocrine therapy is the standard of care for patients with hormone receptor-positive/human epidermal growth factor 2-negative (HR+/HER2-) metastatic breast cancer (mBC), but real-world effectiveness data for patients with lung or liver metastases are limited. This retrospective study included data from the US Flatiron Health database of patients with HR+/HER2- mBC and lung or liver metastases treated with first-line palbociclib (PAL) plus an aromatase inhibitor (AI) or an AI alone in routine clinical practice. Overall survival (OS) and real-world progression-free survival (rwPFS) were assessed. A total of 891 patients were included (622 with lung metastasis, 376 with liver metastasis, and 107 with both lung and liver metastasis). After stabilized inverse probability of treatment weighting to balance patient characteristics, PAL + AI versus AI alone was associated with significantly prolonged OS (HR = 0.62; p < 0.001) and rwPFS (HR = 0.55; p < 0.001) in patients with lung metastases and numerically longer OS (HR = 0.73; p = 0.056) and significantly longer rwPFS (HR = 0.57, p < 0.001) for those with liver metastases. Overall, PAL + AI versus AI alone was associated with prolonged OS and rwPFS in routine clinical practice, supporting the use of first-line PAL + AI for patients with HR+/HER2- mBC with lung and/or liver metastases.

Treatment Patterns and Outcomes Among Patients With Advanced or Recurrent Endometrial Cancer Initiating First-Line Therapy in the United States

Background: Patients with advanced or recurrent endometrial cancer (EC) typically have limited treatment options and poor long-term survival outcomes following first-line therapy. Real-world treatment patterns and survival outcomes data are limited for patients in this setting. Objectives: The objective of this retrospective study was to describe real-world demographics, clinical characteristics, treatment patterns, and overall survival among patients in the United States with primary advanced or recurrent EC who initiated at least 1 line of therapy (LOT). Methods: Patients with a diagnosis of primary advanced or recurrent EC in a real-world database from January 1, 2013, to July 31, 2021, were included. The date for inclusion was the date of EC diagnosis documentation; patients were indexed for treatment patterns and outcomes at the start of the first LOT and at the start of each subsequent LOT they initiated. Data were stratified by subgroups of patients who had mismatch repair deficient (dMMR) or microsatellite instability-high (MSI-H) tumors. Results: A total of 1961 patients who received at least 1 LOT were included. Most patients in this cohort, and the dMMR/MSI-H subgroup, received a platinum combination as first-line treatment, with carboplatin-paclitaxel being the most common regimen. Only 53% of patients who received first-line treatment subsequently received second-line therapy. Of the patients who received at least 1 LOT, use of immunotherapy in the second-line setting was more common in the dMMR/MSI-H subgroup. Median overall survival ranged from 14.1 to 31.8 months across the 5 most frequently used first-line treatment regimens in the ≥1 LOT cohort and became shorter with each subsequent LOT. Discussion: The use of platinum-based chemotherapy for first-line treatment of advanced or recurrent EC predominates in the real-world setting, despite the poor long-term survival outcomes associated with most of these regimens. Conclusions: Patients with recurrent/advanced EC have a poor prognosis, highlighting the need for therapies with more durable benefits.

Evaluation of US oncology electronic health record real-world data to reduce uncertainty in health technology appraisals: a retrospective cohort study

OBJECTIVES

Examine whether data from early access to medicines in the USA can be used to inform National Institute for Health and Care Excellence (NICE) health technology assessments (HTA) in oncology.

DESIGN

Retrospective cohort study.

SETTING

Oncology-based community and academic treatment centres in the USA.

PARTICIPANTS

Patients present in a nationwide electronic health record (EHR)-derived deidentified database.

INTERVENTIONS

Cancer drugs that underwent NICE technology appraisal (TA) between 2014 and 2019.

PRIMARY AND SECONDARY OUTCOME MEASURES

The count and follow-up time of US patients, available in the EHR, who were exposed to cancer drugs of interest in the period between Food and Drug Administration (FDA) approval and dates relevant to the NICE appraisal process.

RESULTS

In 59 of 60 TAs analysed, the cancer therapy was approved in the USA before the final appraisal by NICE. The median time from FDA approval to the publication of NICE recommendations was 18.5 months, at which time the US EHR-derived database had, on average, 269 patients (SD=356) exposed to the new therapy, with a median of 75.3 person-years (IQR: 13.1-173) in time-at-risk. A case study generated evidence on real-world overall survival and treatment duration.

CONCLUSIONS

Across different cancer therapies, there was substantial variability in US real-world data accumulated between FDA approval and NICE decision milestones. The applicability of these data to generate evidence for HTA decision-making should be assessed on a case-by-case basis depending on the intended HTA use case.

Real-World Effectiveness of Palbociclib Plus Aromatase Inhibitors in African American Patients With Metastatic Breast Cancer

BACKGROUND

Disparities in survival and clinical outcomes between African American and White patients with breast cancer (BC) are well documented, but African American patients have not been well represented in randomized clinical trials of CDK4/6 inhibitors. Real-world studies can provide evidence for effective treatment strategies for underreported patient populations.

PATIENTS AND METHODS

This retrospective analysis of African American patients with HR+/HER2- metastatic breast cancer (mBC) from the Flatiron Health longitudinal database evaluated treatments for patients with BC in routine clinical practice in the US. Patients initiated first-line therapy with palbociclib plus an aromatase inhibitor (AI) or AI alone between February 2015 and March 2020. Outcomes assessed included overall survival (OS) and real-world progression-free survival (rwPFS) until September 2020.

RESULTS

Of 270 eligible patients, 127 (median age 64 years) were treated with palbociclib + AI, and 143 (median age 68 years) were treated with an AI. Median follow-up was 24.0 months for palbociclib + AI and 18.2 months for AI-treated patients. Median OS was not reached (NR; 95% CI, 38.2-NR) in the palbociclib + AI group versus 28.2 months (95% CI, 19.2-52.8) in the AI group (adjusted HR, 0.56; 95% CI, 0.36-0.89; P = .013). Median rwPFS was 18.0 months (95% CI, 12.4-26.7) in the palbociclib + AI group and 10.5 months (95% CI, 7.0-13.4) in the AI group (adjusted HR, 0.74; 95% CI, 0.47-1.17; P = .199).

CONCLUSION

This comparative analysis of palbociclib + AI versus AI alone indicates that palbociclib combined with endocrine therapy in the first line is associated with improved effectiveness for African American patients with HR+/HER2- mBC in real-world settings.

TRIAL NUMBER

NCT05361655.

Real-World Treatment Patterns and Outcomes of First-Line Immunotherapy Among Patients With Advanced Nonsquamous NSCLC Harboring BRAF, MET, or HER2 Alterations

Introduction

Data on utilization and clinical outcomes of programmed cell death protein or programmed death-ligand 1 (PD-[L]1) inhibitors in NSCLC with uncommon oncogenic alterations is limited.

Methods

This retrospective study used a deidentified U.S. nationwide clinicogenomic database to select patients with advanced nonsquamous NSCLC without EGFR, ALK, or ROS1 alterations, diagnosed from January 1, 2016 to September 30, 2020, who initiated first-line therapy. Our objectives were to summarize characteristics and treatment patterns for patients with four little-studied genomic alterations or driver-negative NSCLC. We estimated Kaplan-Meier real-world time on treatment (rwTOT) and time to next treatment for patients receiving PD-(L)1 inhibitors. The data cutoff was September 30, 2021.

Results

Of the 3971 eligible patients, 84 (2%) had NSCLC with BRAF V600E mutation, 117 (3%) had MET exon 14 skipping mutation, 130 (3%) had MET amplification, 91 (2%) had ERBB2 activation mutation, and 691 patients (17%) had driver-negative NSCLC. Patient characteristics differed among cohorts as expected. The most common first-line regimen in each cohort was a PD-(L)1 inhibitor as monotherapy or in combination with chemotherapy. The median rwTOT with anti-PD-(L)1 monotherapy was 4.6 months in the driver-negative cohort and ranged from 2.9 months (ERBB2 mutation) to 7.6 months (BRAF V600E mutation). The median rwTOT with anti-PD-(L)1-chemotherapy combination was 5.2 months in the driver-negative cohort and 6 months in all but the BRAF V600E cohort (17.5 mo). The patterns of real-world time to next treatment results were similar.

Conclusions

Substantial use of anti-PD-(L)1 therapy and associated clinical outcomes are consistent with previous real-world findings and suggest no detriment from PD-(L)1 inhibitors for advanced nonsquamous NSCLC harboring one of these four genomic alterations relative to driver-negative NSCLC.

Real-world treatment patterns and effectiveness of palbociclib plus an aromatase inhibitor in patients with metastatic breast cancer aged 75 years or older

Background

Elderly patients are generally underrepresented in oncology clinical trials; therefore, real-world data are needed to inform clinical management of elderly patients with hormone receptor-positive/human epidermal growth factor receptor 2-negative (HR+/HER2-) metastatic breast cancer (mBC). This subanalysis of the P-REALITY X study (NCT05361655) evaluated palbociclib treatment patterns and comparative effectiveness of palbociclib plus an aromatase inhibitor (AI) versus an AI alone among patients with HR+/HER2- mBC aged ≥ 75 years treated in routine clinical practice in the United States.

Methods

This retrospective observational cohort study used electronic health records from the Flatiron Health Analytic Database. Palbociclib treatment patterns, overall survival (OS), real-world progression-free survival (rwPFS), and time to chemotherapy (TTC) were evaluated. Three methods were used for comparative analyses: (1) an unadjusted analysis, (2) stabilized inverse probability treatment weighting (sIPTW; primary analysis), and (3) propensity score matching (PSM; sensitivity analysis).

Results

A total of 961 patients aged ≥ 75 years with HR+/HER2- mBC were identified who started palbociclib plus an AI (n = 313) or an AI alone (n = 648) as first-line (1L) therapy between February 2015 and March 2020 (data cut-off: September 30, 2020). Among patients in the palbociclib plus an AI group with a documented palbociclib starting dose (n = 306), approximately 75% started palbociclib at 125 mg/day, and approximately 40% experienced dose adjustment. After sIPTW, patients treated with palbociclib plus an AI versus an AI alone had significantly improved OS (median of 43.0 vs. 32.4 months; hazard ratio [HR], 0.66 [95% confidence interval (CI), 0.51-0.84]; P = 0.0007), rwPFS (median of 20.0 vs. 15.0 months; HR, 0.72 (0.59-0.89); P = 0.0021), and TTC (median of 40.2 vs. 27.4 months; HR, 0.69 [0.55-0.87]; P = 0.0014). These significant improvements in OS, rwPFS, and TTC remained consistent in the unadjusted analysis and after PSM.

Conclusion

This real-world comparative analysis demonstrated that 1L palbociclib plus an AI is associated with improved effectiveness compared with an AI alone among patients with HR+/HER2- mBC aged ≥ 75 years. These findings support palbociclib plus an AI as a standard-of-care 1L treatment for elderly patients with HR+/HER2- mBC.

A pan-sarcoma landscape of telomeric content shows that alterations in RAD51B and GID4 are associated with higher telomeric content

Tumor cells need to activate a telomere maintenance mechanism, enabling limitless replication. The bulk of evidence supports that sarcomas predominantly use alternative lengthening of telomeres (ALT) mechanism, commonly associated with alterations in ATRX and DAXX. In our dataset, only 12.3% of sarcomas harbored alterations in these genes. Thus, we checked for the presence of other genomic determinants of high telomeric content in sarcomas. Our dataset consisted of 13555 sarcoma samples, sequenced as a part of routine clinical care on the FoundationOne®Heme platform. We observed a median telomeric content of 622.3 telomeric reads per GC-matched million reads (TRPM) across all samples. In agreement with previous studies, telomeric content was significantly higher in ATRX altered and POT1 altered sarcomas. We further observed that sarcomas with alterations in RAD51B or GID4 were enriched in samples with high telomeric content, specifically within uterus leiomyosarcoma for RAD51B and soft tissue sarcoma (not otherwise specified, nos) for GID4, Furthermore, RAD51B and POT1 alterations were mutually exclusive with ATRX and DAXX alterations, suggestive of functional redundancy. Our results propose a role played by RAD51B and GID4 in telomere elongation in sarcomas and open research opportunities for agents aimed at targeting this critical pathway in tumorigenesis.

Association Between Homologous Recombination Repair Biomarkers and Survival in Patients With Solid Tumors

PURPOSE

Mutations in BRCA1 and/or BRCA2 (BRCAm), other homologous recombination repair genes (HRRm), and homologous recombination deficiency (HRD) lead to an accumulation of genomic alterations that can drive tumorigenesis. The prognostic impact of these HRR pathway defects on overall survival (OS) in patients not receiving poly (ADP-ribose) polymerase inhibitors (PARPi) or immunotherapy is unclear. We evaluated the association of HRR biomarkers with OS in patients with advanced solid tumors receiving therapy excluding PARPi and immunotherapy.

METHODS

Deidentified data were collected through December 31, 2020, from a real-world clinicogenomic database (CGDB) with data originating from approximately 280 cancer clinics in the United States. Patients age 18 years and older with an advanced/metastatic diagnosis between 2018 and 2019 for 1 of 15 solid tumors and available data in the CGDB were included. The primary analysis evaluated the association between HRR pathway biomarkers and OS, using start of second-line therapy as the index date (to reduce immortal time bias).

RESULTS

A total of 9,457 patients had available data for BRCA/HRR and 5,792 for HRD status; 4,890 (51.7%) were women and mean (SD) age was 65.9 (11.5) years. For the primary analysis, adjusted hazard ratios for OS were BRCAm (n = 156) versus BRCA wild-type (wt; n = 3,131; 0.83 [95% CI, 0.60 to 1.17]); for HRRm (n = 467) versus HRRwt (n = 282; 0.95 [95% CI, 0.79 to 1.14]); and for HRD-positive (n = 447) versus -negative (n = 1,687; 1.22 [95% CI, 1.02 to 1.47]). Results were similar using start of first-line and start of third-line therapy as index dates.

CONCLUSION

This large, real-world study found no association between OS and either BRCA or HRR status but identified a possible linkage between HRD positivity and shorter median OS in patients with advanced solid tumors who did not receive PARPi or immunotherapy.

Improving Real-World Mortality Data Quality in Oncology Research: Augmenting Electronic Medical Records With Obituary, Social Security Death Index, and Commercial Claims Data

PURPOSE

This study evaluated the relative improvements in mortality data capture of adding different external data to enriched electronic medical records (EMRs) for patients with melanoma.

METHODS

An enriched EMR database, containing structured and unstructured data, was used to evaluate the incremental mortality data capture of the following external data sources: Social Security Administration (SSA), public obituary, and an administrative open-claims database for the claims data set. Overall survival (OS) was assessed for each data set and the composite data set using the Kaplan-Meier method.

RESULTS

A total of 3,882 patients were included in the study. The enriched EMR data set identified 1,085 patients with a death record. The SSA data set identified 213 patients (73 unique when combined with enriched EMR) with a death record, while the obituary data set identified 1,127 patients (241 unique). The administrative claims data set identified 378 patients (73 unique) with a death record; however, all these unique patients were already accounted for in the combined SSA and obituary data set. The composite data set yielded a median OS of 13.39 years, about 4 years shorter than the enriched EMR data set alone (17.63 years).

CONCLUSION

When the enriched EMR data set was augmented with one external data set, the obituary data set provided the most additional value, followed by claims, and then SSA. The augmentation of all the data sources had a significant impact on the OS results compared with enriched EMR alone.

Socioeconomic status and inequities in treatment initiation and survival among patients with cancer, 2011-2022

BACKGROUND

Lower neighborhood socioeconomic status (SES) is associated with suboptimal cancer care and reduced survival. Most studies examining cancer inequities across area-level socioeconomic status tend to use less granular or unidimensional measures and pre-date the COVID-19 pandemic. Here, we examined the association of area-level socioeconomic status on real-world treatment initiation and overall survival among adults with 20 common cancers.

METHODS

This retrospective cohort study used electronic health record-derived deidentified data (Flatiron Health Research Database, 2011-2022) linked to US Census Bureau data from the American Community Survey (2015-2019). Area-level socioeconomic status quintiles (based on a measure incorporating income, home values, rental costs, poverty, blue-collar employment, unemployment, and education information) were computed from the US population and applied to patients based on their mailing address. Associations were examined using Cox proportional hazards models adjusted for diagnosis year, age, sex, performance status, stage, and cancer type.

RESULTS

This cohort included 291 419 patients (47.7% female; median age = 68 years). Patients from low-SES areas were younger and more likely to be Black (21.9% vs 3.3%) or Latinx (8.4% vs 3.0%) than those in high-SES areas. Living in low-SES areas (vs high) was associated with lower treatment rates (hazard ratio = 0.94 [95% confidence interval = 0.93 to 0.95]) and reduced survival (median real-world overall survival = 21.4 vs 29.5 months, hazard ratio = 1.20 [95% confidence interval = 1.18 to 1.22]). Treatment and survival inequities were observed in 9 and 19 cancer types, respectively. Area-level socioeconomic inequities in treatment and survival remained statistically significant in the COVID-19 era (after March 2020).

CONCLUSION

To reduce inequities in cancer outcomes, efforts that target marginalized, low-socioeconomic status neighborhoods are necessary.

Tumor Mutational Burden in Real-World Patients With Pancreatic Cancer: Genomic Alterations and Predictive Value for Immune Checkpoint Inhibitor Effectiveness

PURPOSE

Pancreatic ductal adenocarcinoma (PDAC) is largely considered a nonimmunogenic malignancy; however, approximately 1%, of patients may have tumors with deficient mismatch repair, high microsatellite instability, or high tumor mutational burden (TMB ≥10 mutations/Mb), which may be predictive of response to immune checkpoint inhibitor (ICI) therapy. We sought to analyze outcomes of patients with high-TMB and pathogenic genomic alterations observed in this population.

METHODS

This study included patients with PDAC who underwent comprehensive genomic profiling (CGP) at Foundation Medicine (Cambridge, MA). Clinical data were obtained from a US-wide real-world clinicogenomic pancreatic database. We report genomic alterations in those with high and low TMB, and compare outcomes on the basis of receipt of single-agent ICI or therapy regimens not containing ICI.

RESULTS

We evaluated 21,932 patients with PDAC who had tissue CGP data available, including 21,639 (98.7%) with low-TMB and 293 (1.3%) with high-TMB. Among patients with high-TMB, a greater number of alterations were observed in BRCA2, BRAF, PALB2, and genes of the mismatch repair pathway, whereas fewer alterations were observed in KRAS. Among patients who received an ICI (n = 51), those with high-TMB had more favorable median overall survival when compared with the low-TMB subset (25.7 v 5.2 months; hazard ratio, 0.32; 95% CI, 0.11 to 0.91; P = .034).

CONCLUSION

Longer survival was observed in patients with high-TMB receiving ICI compared with those with low-TMB. This supports the role of high-TMB as a predictive biomarker for efficacy of ICI therapy in PDAC. Additionally, we report higher rates of BRAF and BRCA2 mutations and lower rates of KRAS mutation among patients with PDAC and high-TMB, which to our knowledge, is a novel finding.

Inverse Probability of Treatment Weighting and Confounder Missingness in Electronic Health Record-based Analyses: A Comparison of Approaches Using Plasmode Simulation

BACKGROUND

Electronic health record (EHR) data represent a critical resource for comparative effectiveness research, allowing investigators to study intervention effects in real-world settings with large patient samples. However, high levels of missingness in confounder variables is common, challenging the perceived validity of EHR-based investigations.

METHODS

We investigated performance of multiple imputation and propensity score (PS) calibration when conducting inverse probability of treatment weights (IPTW)-based comparative effectiveness research using EHR data with missingness in confounder variables and outcome misclassification. Our motivating example compared effectiveness of immunotherapy versus chemotherapy treatment of advanced bladder cancer with missingness in a key prognostic variable. We captured complexity in EHR data structures using a plasmode simulation approach to spike investigator-defined effects into resamples of a cohort of 4361 patients from a nationwide deidentified EHR-derived database. We characterized statistical properties of IPTW hazard ratio estimates when using multiple imputation or PS calibration missingness approaches.

RESULTS

Multiple imputation and PS calibration performed similarly, maintaining ≤0.05 absolute bias in the marginal hazard ratio even when ≥50% of subjects had missing at random or missing not at random confounder data. Multiple imputation required greater computational resources, taking nearly 40 times as long as PS calibration to complete. Outcome misclassification minimally increased bias of both methods.

CONCLUSION

Our results support multiple imputation and PS calibration approaches to missingness in missing completely at random or missing at random confounder variables in EHR-based IPTW comparative effectiveness analyses, even with missingness ≥50%. PS calibration represents a computationally efficient alternative to multiple imputation.

A real-world data analysis of predictors of early mortality after a diagnosis of multiple myeloma

BACKGROUND

Despite the increased availability and use of novel therapies for multiple myeloma, early mortality is a pervasive challenge with a significant impact on older adults. Reported rates and predictors of early mortality have varied in the literature, with most studies seldom focusing on community-treated patients.

METHODS

In this retrospective cohort analysis of a real-world electronic health record-derived deidentified database of 7512 patients newly diagnosed with multiple myeloma between January 1, 2011, and February 2, 2021, and treated primarily in US-based community oncology practices, factors associated with early mortality (defined as death within 6 months after the multiple myeloma diagnosis) were examined with the use of binary logistic regression.

RESULTS

The median age was 70 years overall. We found an overall early mortality rate of 8.3%, with 73% of early deaths occurring in those aged ≥70 years. Among the early deaths, only 49 patients (8.7%) had documented disease progression before death (median time to progression, 30 days [interquartile range, 7-53 days]). Baseline factors associated with higher odds of early mortality included an Eastern Cooperative Oncology Group performance status (ECOG PS) ≥ 2, Revised International Staging System (R-ISS) stage III, an age ≥ 70 years, receipt of proteasome inhibitor-doublet therapy, a light-chain isotype, and the presence of renal dysfunction (estimated glomerular filtration rate < 30 mL/min). Among those aged ≥70 years, ECOG PS ≥ 2 and R-ISS stage III remained the strongest predictors of early mortality.

CONCLUSIONS

Early mortality disproportionately affects older adults (aged ≥70 years) with multiple myeloma. Interventions to support this population are needed to reduce disparate survival outcomes.

PLAIN LANGUAGE SUMMARY

Factors associated with an increased risk of dying within 6 months (early mortality) of a new diagnosis of multiple myeloma (MM) among 7512 mostly community-treated patients with MM were evaluated. The early mortality rate was 8.3%; among those deaths, 49 patients (8.7%) had documented evidence of MM progression before death. The risk of early mortality was greatest for older patients (aged ≥70 years) and those with a poor performance status, poor kidney function, a higher disease stage, and light-chain MM and those receiving two-drug MM therapies. These findings highlight the need for supportive interventions geared toward older adults with MM.

Real-World First-Line Use of Pertuzumab With Different Taxanes for Human Epidermal Growth Factor Receptor 2-Positive Metastatic Breast Cancer: A Comparative Effectiveness Study Using US Electronic Health Records

PURPOSE

On the basis of the results from CLEOPATRA, pertuzumab plus trastuzumab and chemotherapy is the first-line standard of care for human epidermal growth factor receptor 2 (HER2)-positive metastatic breast cancer (MBC). However, discrepancies have been reported between clinical trial and real-world outcomes. We report real-world outcomes for patients with HER2-positive MBC treated with first-line pertuzumab plus trastuzumab and a taxane in routine clinical practice in the United States.

METHODS

A retrospective analysis was conducted using electronic health record-derived deidentified data from the Flatiron Health database. Patients were grouped according to the first taxane received (paclitaxel/nab-paclitaxel or docetaxel). Median real-world progression-free survival (rwPFS) and overall survival (rwOS) was estimated using Kaplan-Meier methodology. Subgroup analyses were conducted in patients treated with docetaxel who met CLEOPATRA's key eligibility criteria.

RESULTS

We included 1,065 patients; 313 patients received paclitaxel/nab-paclitaxel and 752 received docetaxel. Patients who received paclitaxel/nab-paclitaxel were older, had a worse Eastern Cooperative Oncology Group Performance Status, and had more recurrent metastatic disease compared with the docetaxel group. After adjustment for potential confounders, similar median rwPFS (inverse probability of treatment weighted average treatment effect for the treated [IPTW-ATT] hazard ratio [HR], 1.09; 95% CI, 0.9 to 1.3; P = .365) and rwOS (IPTW-ATT HR, 1.23; 95% CI, 0.96 to 1.58; P = .101) was observed between treatment groups. In the subgroup of CLEOPATRA-eligible patients, median rwPFS and rwOS were 16.9 months and 57.8 months, respectively.

CONCLUSION

There was no statistically significant difference in real-world outcomes between patients treated with paclitaxel/nab-paclitaxel and those treated with docetaxel. Selecting patients using key CLEOPATRA eligibility criteria resulted in rwPFS and rwOS similar to those observed in CLEOPATRA, highlighting the importance of ensuring similar patient populations when comparing clinical trial and real-world data.

Real-world treatment patterns and clinical outcomes with first-line therapy in patients with locally advanced/metastatic urothelial carcinoma by cisplatin-eligibility

INTRODUCTION

Patients with locally advanced/metastatic urothelial carcinoma (la/mUC) have a poor prognosis. With recent therapeutic advances, data on real-world treatment patterns and overall survival (OS) in patients with la/mUC treated with first-line therapy are limited, particularly when comparing patients who are cisplatin-ineligible versus cisplatin-eligible.

METHODS

This was a retrospective observational study of real-world first-line treatment patterns and OS in patients with la/mUC stratified by cisplatin-eligibility and treatment. Data were from a nationwide electronic health record-derived de-identified database. Eligible patients were adults diagnosed with la/mUC from May 2016 to April 2021 and followed until death or end of data availability in January 2022. OS stratified by first-line treatment and cisplatin eligibility was estimated using Kaplan-Meier methods and compared via multivariable Cox proportional-hazard models adjusted for clinical covariates.

RESULTS

Of 4,757 patients with la/mUC, 3,632 (76.4%) received first-line treatment, with 2,029 (55.9%) cisplatin-ineligible and 1,603 (44.1%) cisplatin-eligible. Patients who were cisplatin-ineligible were older (mean age, 74.9 vs. 68.8 years) and had lower CrCl (median, 46.4 vs. 87.0 ml/min). Only 43.8% of patients receiving first-line treatment (37.6% cisplatin-ineligible vs. 51.6% cisplatin-eligible) received second-line therapy. Median OS in all patients receiving first-line treatment was 10.8 (95% CI, 10.2-11.3) months and was shorter in patients who were cisplatin-ineligible than cisplatin-eligible (8.5 [95% CI, 7.8-9.0] vs. 14.4 [13.3-16.1]; hazard ratio [HR], 0.9 [0.7-1.1]). Cisplatin-based therapy was associated with longer OS (17.6 [15.1-20.4] months) than other first-line treatments (the shortest OS was with PD-1/L1 inhibitor monotherapy; 7.7 [6.8-8.8] months), including among patients who were classified as cisplatin-ineligible.

CONCLUSIONS

Outcomes for patients with newly diagnosed la/mUC are poor, particularly for patients who are cisplatin-ineligible and/or do not receive cisplatin-based therapy. Many patients with la/mUC did not receive first-line treatment and among those who did, fewer than half received second-line therapy. These data highlight the need for more effective first-line therapies for all patients with la/mUC.

Clinical characteristics and overall survival among acute myeloid leukemia patients with TP53 gene mutation or chromosome 17p deletion

Approximately 5% to 15% of acute myeloid leukemia (AML) patients have TP53 gene mutations (TP53m), which are associated with very poor outcomes. Adults (≥18 years) with a new AML diagnosis were included from a nationwide, de-identified, real-world database. Patients receiving first-line therapy were divided into three cohorts: venetoclax (VEN) + hypomethylating agents (HMAs; Cohort A), intensive chemotherapy (Cohort B), or HMA without VEN (Cohort C). A total of 370 newly diagnosed AML patients with TP53m (n = 124), chromosome 17p deletion (n = 166), or both (n = 80) were included. The median age was 72 years (range, 24-84); most were male (59%) and White (69%). Baseline bone marrow (BM) blasts were ≤30%, 31%-50%, and >50% in 41%, 24%, and 29% of patients in Cohorts A, B, and C, respectively. BM remission (<5% blasts) with first-line therapy was reported in 54% of patients (115/215) overall, and 67% (38/57), 62% (68/110), and 19% (9/48) for respective cohorts (median BM remission duration: 6.3, 6.9, and 5.4 months). Median overall survival (95% CI) was 7.4 months (6.0-8.8) for Cohort A, 9.4 months (7.2-10.4) for Cohort B, and 5.9 months (4.3-7.5) for Cohort C. There were no differences in survival by treatment type after adjusting for the effects of relevant covariates (Cohort A vs. C adjusted hazard ratio [aHR] = 0.9; 95% CI, 0.7-1.3; Cohort A vs. B aHR = 1.0; 95% CI, 0.7-1.5; and Cohort C vs. B aHR = 1.1; 95% CI, 0.8-1.6). Patients with TP53m AML have dismal outcomes with current therapies, demonstrating the high unmet need for improved treatments.

Postprediction Inference for Clinical Characteristics Extracted With Machine Learning on Electronic Health Records

PURPOSE

Real-world data (RWD) derived from electronic health records (EHRs) are often used to understand population-level relationships between patient characteristics and cancer outcomes. Machine learning (ML) methods enable researchers to extract characteristics from unstructured clinical notes, and represent a more cost-effective and scalable approach than manual expert abstraction. These extracted data are then used in epidemiologic or statistical models as if they were abstracted observations. Analytical results derived from extracted data in this way may differ from those given by abstracted data, and the magnitude of this difference is not directly informed by standard ML performance metrics.

METHODS

In this paper, we define the task of postprediction inference, which is to recover similar estimation and inference from an ML-extracted variable that would be obtained from abstracting the variable. We consider fitting a Cox proportional hazards model that uses a binary ML-extracted variable as a covariate and evaluate four approaches for postprediction inference in this setting. The first two approaches only require the ML-predicted probability, while the latter two additionally require a labeled (human abstracted) validation data set.

RESULTS

Our results for both simulated data and EHR-derived RWD from a national cohort demonstrate that we can improve inference from ML-extracted variables by leveraging a limited amount of labeled data.

CONCLUSION

We describe and evaluate methods for fitting statistical models using ML-extracted variables subject to model error. We show that estimation and inference is generally valid when using extracted data from high-performing ML models. More complex methods that incorporate auxiliary labeled data provide further improvements.

Chemoimmunotherapy vs. Immunotherapy for First Line Treatment of Advanced Non-small Cell Lung Cancer With a PD-L1 Expression ≥50% or ≥90

BACKGROUND

Evidence about the comparative effectiveness of chemoimmunotherapy vs. immunotherapy alone in patients with advanced non-small cell lung cancer (aNSCLC) and high PD-L1 expression (≥50%) or very high PD-L1 expression (≥90%) is limited because of the lack of head-to-head clinical trials.

OBJECTIVE

To compare survival in aNSCLC patients receiving first-line chemoimmunotherapy vs. immunotherapy in both the PD-L1 expression ≥50% or ≥90% subgroups, accounting for potential confounders that may influence physician decision-making.

METHODS

This cohort study used a nationwide electronic health record derived database to identify newly diagnosed cases of aNSCLC patients with PD-L1 expression of ≥50% who initiated first-line systemic therapy between October 2016 and October 2021. The exposure of interest was first-line therapy with chemoimmunotherapy or immunotherapy among patients with PD-L1 expression ≥50% or ≥90%. Survival was assessed using Kaplan-Meier curves and Cox regression. Propensity score-based inverse probability of weighting (IPW) was used to control for confounding. Because of nonproportionality of hazards, we estimated hazard ratios over the first 6 months and after 6 months for the overall cohort, and over the first 12 months and after 12 months for a subgroup of persons with a PD-L1 expression ≥90%.

RESULTS

We identified 3086 subjects who met inclusion criteria, of whom 32% received chemoimmunotherapy and 68% received immunotherapy alone. Chemoimmunotherapy was associated with no survival advantage vs. immunotherapy alone during the entire follow-up period (IPW-adjusted Hazard Ratio [aHR] 0.98, 95% CI, 0.86-1.12), but was associated with a survival benefit during the first 6 months (aHR 0.74, 95% CI, 0.61-0.90). Similarly, in the subgroup of patients with a PD-L1 expression ≥90%, chemoimmunotherapy was associated with no overall survival advantage during the entire follow-up period (aHR 0.99, 95% CI, 0.87-1.22), but was associated with a survival benefit during the first 12 months (aHR 0.74, 95% CI, 0.57-0.97).

CONCLUSION

Chemoimmunotherapy was not associated with an overall benefit over immunotherapy alone, although was associated with an early survival advantage in both the overall cohort and the subgroup of patients with a PD-L1 expression ≥90%. Future studies should focus on identifying the characteristics of higher risk patients that may benefit from the addition of chemotherapy.

Data Challenges for Externally Controlled Trials: Viewpoint

The preferred evidence of a large randomized controlled trial is difficult to adopt in scenarios, such as rare conditions or clinical subgroups with high unmet needs, and evidence from external sources, including real-world data, is being increasingly considered by decision makers. Real-world data originate from many sources, and identifying suitable real-world data that can be used to contextualize a single-arm trial, as an external control arm, has several challenges. In this viewpoint article, we provide an overview of the technical challenges raised by regulatory and health reimbursement agencies when evaluating comparative efficacy, such as identification, outcome, and time selection challenges. By breaking down these challenges, we provide practical solutions for researchers to consider through the approaches of detailed planning, collection, and record linkage to analyze external data for comparative efficacy.

Mismatch repair deficiency, next-generation sequencing-based microsatellite instability, and tumor mutational burden as predictive biomarkers for immune checkpoint inhibitor effectiveness in frontline treatment of advanced stage endometrial cancer

OBJECTIVE

Molecular profiling is developing to inform treatment in endometrial cancer. Using real world evidence, we sought to evaluate frontline immune checkpoint inhibitor vs chemotherapy effectiveness in advanced endometrial cancer, stratified by Tumor Mutational Burden (TMB) ≥10 mut/MB and microsatellite instability (MSI).

METHODS

Patients with advanced endometrial cancer in the US-based de-identified Flatiron Health-Foundation Medicine Clinico-Genomic Database were included. Data originated from patients treated between January 2011- March 2022 at 280 US clinics. Next-generation sequencing assays were performed via FoundationOne or FoundationOneCDx. Longitudinal clinical data were derived from electronic health records. Immune checkpoint inhibitor treatment included pembrolizumab, dostarlimab, and nivolumab monotherapies. Time to next treatment, time to treatment discontinuation, and overall survival were assessed with the log-rank test and Cox proportional hazard models with adjusted hazard ratios (aHR) for known prognostic factors. We used the Likelihood ratio test to compare biomarker performance.

RESULTS

A total of 343 patients received chemotherapy and 28 received immune checkpoint inhibitor monotherapy as frontline treatment. Patients who received monotherapy were more likely to be stage III at diagnosis (immune checkpoint inhibitor: 54.6% vs chemotherapy: 15.0%; p<0.001) and more likely to test MSI-high via next-generation sequencing (immune checkpoint inhibitor: 53.6% vs chemotherapy: 19.2%; p<0.001). In MSI-high cancers, single-agent immune checkpoint inhibitor had a more favorable time to next treatment (aHR: 0.18, p=0.001) and overall survival (aHR 0.29, p=0.045). Additional analyses on 70 unique tumor specimens revealed mismatch repair deficiency (dMMR) via immunohistochemistry and MSI-high via next-generation sequencing concordance (91%), with nominal improvement of MSI over dMMR to predict time to treatment discontinuation (p=0.030), time to next treatment (p=0.032), and overall survival (p=0.22). MSI status was concordant with tumor mutational burden ≥10 in 94.3% of cases.

CONCLUSION

Immune checkpoint inhibitors may have improved efficacy over chemotherapy in frontline treatment for advanced endometrial cancer defined by MSI-high using next-generation sequencing as a nominally better predictor of outcomes than dMMR with immunohistochemistry. This provides the biologic rationale of active phase III trials.

Replication of Real-World Evidence in Oncology Using Electronic Health Record Data Extracted by Machine Learning

Meaningful real-world evidence (RWE) generation requires unstructured data found in electronic health records (EHRs) which are often missing from administrative claims; however, obtaining relevant data from unstructured EHR sources is resource-intensive. In response, researchers are using natural language processing (NLP) with machine learning (ML) techniques (i.e., ML extraction) to extract real-world data (RWD) at scale. This study assessed the quality and fitness-for-use of EHR-derived oncology data curated using NLP with ML as compared to the reference standard of expert abstraction. Using a sample of 186,313 patients with lung cancer from a nationwide EHR-derived de-identified database, we performed a series of replication analyses demonstrating some common analyses conducted in retrospective observational research with complex EHR-derived data to generate evidence. Eligible patients were selected into biomarker- and treatment-defined cohorts, first with expert-abstracted then with ML-extracted data. We utilized the biomarker- and treatment-defined cohorts to perform analyses related to biomarker-associated survival and treatment comparative effectiveness, respectively. Across all analyses, the results differed by less than 8% between the data curation methods, and similar conclusions were reached. These results highlight that high-performance ML-extracted variables trained on expert-abstracted data can achieve similar results as when using abstracted data, unlocking the ability to perform oncology research at scale.

Outcomes of hepatocellular carcinoma by etiology with first-line atezolizumab and bevacizumab: a real-world analysis

PURPOSE

Hepatocellular carcinoma (HCC) is a common and deadly form of liver cancer. Combination atezolizumab and bevacizumab has improved the outcomes for patients with advanced disease. We sought to determine the impact of etiology on outcomes of patients treated with atezolizumab and bevacizumab.

METHODS

This study used a real-world database. The primary outcome was overall survival (OS) by etiology of HCC; the secondary outcome was real-world time to treatment discontinuation (rwTTD). Time-to-event analyses was performed by the Kaplan-Meier method; the log-rank test to assess for differences by etiology from date of first receipt of atezolizumab and bevacizumab. The Cox proportional hazards model was used to calculate hazard ratios.

RESULTS

In total, 429 patients were included (n = 216 Viral-HCC; n = 68 Alcohol-HCC; n = 145, NASH-HCC). The median overall survival for the entire cohort was 9.4 months (95% CI 7.1-10.9). Compared with Viral-HCC, the hazard ratio (HR) of death was 1.11 (95% CI 0.74-1.68, p = 0.62) for Alcohol-HCC and was 1.34 (95% CI 0.96-1.86, p = 0.08) for NASH-HCC. The median rwTTD for the entire cohort was 5.7 months (95% CI 5.0-7.0 months). The HR of rwTTD was 1.24 (95% CI 0.86-1.77, p = 0.25) for Alcohol-HCC and was 1.31 (95% CI 0.98-1.75, p = 0.06) in reference to TTD with Viral-HCC.

CONCLUSIONS

In this real-world cohort of patients with HCC receiving first-line atezolizumab and bevacizumab, we did not identify an association between etiology and OS or rwTTD. This suggests that the efficacy of atezolizumab and bevacizumab may be similar across HCC etiologies. Further prospective studies are needed to confirm these findings.