Real-world and clinical trial outcomes in large B-cell lymphoma with axicabtagene ciloleucel across race and ethnicity.

ABSTRACT: Axicabtagene ciloleucel (axi-cel) is an autologous anti-CD19 chimeric antigen receptor (CAR) T-cell therapy approved for relapsed/refractory (R/R) large B-cell lymphoma (LBCL). Despite extensive data supporting its use, outcomes stratified by race and ethnicity groups are limited. Here, we report clinical outcomes with axi-cel in patients with R/R LBCL by race and ethnicity in both real-world and clinical trial settings. In the real-world setting, 1290 patients who received axi-cel between 2017 and 2020 were identified from the Center for International Blood and Marrow Transplant Research database; 106 and 169 patients were included from the ZUMA-1 and ZUMA-7 trials, respectively. Overall survival was consistent across race/ethnicity groups. However, non-Hispanic (NH) Black patients had lower overall response rate (OR, 0.37; 95% CI, 0.22-0.63) and lower complete response rate (OR, 0.57; 95% CI, 0.33-0.97) than NH White patients. NH Black patients also had a shorter progression-free survival vs NH White (HR, 1.41; 95% CI, 1.04-1.90) and NH Asian patients (HR, 1.67; 95% CI, 1.08-2.59). NH Asian patients had a longer duration of response than NH White (HR, 0.56; 95% CI, 0.33-0.94) and Hispanic patients (HR, 0.54; 95% CI, 0.30-0.97). There was no difference in cytokine release syndrome by race/ethnicity; however, higher rates of any-grade immune effector cell-associated neurotoxicity syndrome were observed in NH White patients than in other patients. These results provide important context when treating patients with R/R LBCL with CAR T-cell therapy across different racial and ethnic groups. ZUMA-1 and ZUMA-7 (ClinicalTrials.gov identifiers: #NCT02348216 and #NCT03391466, respectively) are registered on ClinicalTrials.gov.

Electronic phenotyping of health outcomes of interest using a linked claims-electronic health record database: Findings from a machine learning pilot project.

OBJECTIVE: Claims-based algorithms are used in the Food and Drug Administration Sentinel Active Risk Identification and Analysis System to identify occurrences of health outcomes of interest (HOIs) for medical product safety assessment. This project aimed to apply machine learning classification techniques to demonstrate the feasibility of developing a claims-based algorithm to predict an HOI in structured electronic health record (EHR) data. MATERIALS AND METHODS: We used the 2015-2019 IBM MarketScan Explorys Claims-EMR Data Set, linking administrative claims and EHR data at the patient level. We focused on a single HOI, rhabdomyolysis, defined by EHR laboratory test results. Using claims-based predictors, we applied machine learning techniques to predict the HOI: logistic regression, LASSO (least absolute shrinkage and selection operator), random forests, support vector machines, artificial neural nets, and an ensemble method (Super Learner). RESULTS: The study cohort included 32 956 patients and 39 499 encounters. Model performance (positive predictive value [PPV], sensitivity, specificity, area under the receiver-operating characteristic curve) varied considerably across techniques. The area under the receiver-operating characteristic curve exceeded 0.80 in most model variations. DISCUSSION: For the main Food and Drug Administration use case of assessing risk of rhabdomyolysis after drug use, a model with a high PPV is typically preferred. The Super Learner ensemble model without adjustment for class imbalance achieved a PPV of 75.6%, substantially better than a previously used human expert-developed model (PPV = 44.0%). CONCLUSIONS: It is feasible to use machine learning methods to predict an EHR-derived HOI with claims-based predictors. Modeling strategies can be adapted for intended uses, including surveillance, identification of cases for chart review, and outcomes research.

A General Propensity Score for Signal Identification Using Tree-Based Scan Statistics.

The tree-based scan statistic (TreeScan; Martin Kulldorff, Harvard Medical School, Boston, Massachusetts) is a data-mining method that adjusts for multiple testing of correlated hypotheses when screening thousands of potential adverse events for signal identification. Simulation has demonstrated the promise of TreeScan with a propensity score (PS)-matched cohort design. However, it is unclear which variables to include in a PS for applied signal identification studies to simultaneously adjust for confounding across potential outcomes. We selected 4 pairs of medications with well-understood safety profiles. For each pair, we evaluated 5 candidate PSs with different combinations of 1) predefined general covariates (comorbidity, frailty, utilization), 2) empirically selected (data-driven) covariates, and 3) covariates tailored to the drug pair. For each pair, statistical alerting patterns were similar with alternative PSs (≤11 alerts in 7,996 outcomes scanned). Inclusion of covariates tailored to exposure did not appreciably affect screening results. Inclusion of empirically selected covariates can provide better proxy coverage for confounders but can also decrease statistical power. Unlike tailored covariates, empirical and predefined general covariates can be applied "out of the box" for signal identification. The choice of PS depends on the level of concern about residual confounding versus loss of power. Potential signals should be followed by pharmacoepidemiologic assessment where confounding control is tailored to the specific outcome(s) under investigation.

FDA Initiative for Drug Facts Label for Over-the-Counter Naloxone.

BACKGROUND: The opioid crisis highlights the need to increase access to naloxone, possibly through regulatory approval for over-the-counter sales. To address industry-perceived barriers to such access, the Food and Drug Administration (FDA) developed a model drug facts label for such sales to assess whether consumers understood the key statements for safe and effective use. METHODS: In this label-comprehension study, we conducted individual structured interviews with 710 adults and adolescents, including 430 adults who use opioids and their family and friends. Eight primary end points were developed to assess user comprehension of each of the key steps in the label. Each of these end points included a prespecified target threshold ranging from 80 to 90% that was evaluated through a comparison of the lower boundary of the 95% exact confidence interval. RESULTS: The results for performance on six primary end points met or exceeded thresholds, including the steps "Check for a suspected overdose" (threshold, 85%; point estimate [PE], 95.8%; 95% confidence interval [CI], 94.0 to 97.1) and "Give the first dose" (threshold, 85%; PE, 98.2%; 95% CI, 96.9 to 99.0). The lower boundaries for four other primary end points ranged from 88.8 to 94.0%. One exception was comprehension of "Call 911 immediately," but this instruction closely approximated the target of 90% (PE, 90.3%; 95% CI, 87.9 to 92.4). Another exception was comprehension of the composite step of "Check, give, and call 911 immediately" (threshold, 85%; PE, 81.1%; 95% CI, 78.0 to 83.9). CONCLUSIONS: Consumers met thresholds for sufficient understanding of six of eight components of the instructions in the drug facts label for naloxone use and came close on two others. Overall, the FDA found that the model label was adequate for use in the development of a naloxone product intended for over-the-counter sales.

Acute Kidney Injury During Treatment with Intravenous Acyclovir for Suspected or Confirmed Neonatal Herpes Simplex Virus Infection.

OBJECTIVE: To describe the epidemiology of and risk factors associated with acute kidney injury (AKI) during acyclovir treatment in neonates and infants. STUDY DESIGN: We conducted a multicenter (n = 4), retrospective cohort study of all hospitalized infants age <60 days treated with intravenous acyclovir (≥1 dose) for suspected or confirmed neonatal herpes simplex virus disease from January 2011 to December 2015. Infants with serum creatinine measured both before acyclovir (baseline) and during treatment were included. We classified AKI based on changes in creatinine according to published neonatal AKI criteria and performed Cox regression analysis to evaluate risk factors for AKI during acyclovir treatment. RESULTS: We included 1017 infants. The majority received short courses of acyclovir (median, 5 doses). Fifty-seven infants (5.6%) developed AKI during acyclovir treatment, with an incidence rate of AKI at 11.6 per 1000 acyclovir days. Cox regression analysis identified having confirmed herpes simplex virus disease (OR, 4.35; P = .002), receipt of ≥2 concomitant nephrotoxic medications (OR, 3.07; P = .004), receipt of mechanical ventilation (OR, 5.97; P = .001), and admission to an intensive care unit (OR, 6.02; P = .006) as risk factors for AKI during acyclovir treatment. CONCLUSIONS: Among our cohort of infants exposed to acyclovir, the rate of AKI was low. Sicker infants and those exposed to additional nephrotoxic medications seem to be at greater risk for acyclovir-induced toxicity and warrant closer monitoring.

Performance evaluation of regression splines for propensity score adjustment in post-market safety analysis with multiple treatments.

Observational studies provide a core resource in assessing post-market drug safety and effectiveness. Propensity scores are a predominant method for confounding adjustment to achieve unbiased estimation of average treatment effects in observational data. However, the use of propensity score methods has been limited to comparing two treatment groups, while medical situations frequently present with multiple treatment options. Inverse probability of treatment weighting (IPTW) is a popular propensity score adjustment method, but its performance degrades with decreased positivity leading to extreme weights, a problem that can be amplified with multiple treatment groups. Meanwhile, regression on a spline of the propensity score has shown favorable performance compared to other propensity score methods in recent studies involving two treatments. This project utilizes a simulation study to compare IPTW and propensity score splines as adjustment methods in a three-treatment setting. We test a variety of spline methods, including natural cubic splines with varying numbers of interior knots, and thin-plate regression splines. We vary several parameters across simulations, including the degree of propensity score overlap among treatment groups, treatment prevalence, outcome prevalence, and true marginal relative risk. We assess methods based on their bias, root mean squared error, and coverage of the true marginal relative risk across simulations. We find that all methods perform similarly well when there is good propensity score distribution overlap. However, with even moderate decrease in overlap or low outcome prevalence, IPTW produces more biased estimates and higher variance than propensity score splines. Low treatment prevalence or unequal treatment prevalences across groups also worsens IPTW performance. Overall, a natural cubic spline with a relatively small number of interior knots provides good performance across a range of simulations.

Comparative Stroke, Bleeding, and Mortality Risks in Older Medicare Patients Treated with Oral Anticoagulants for Nonvalvular Atrial Fibrillation.

BACKGROUND: Nonvitamin K antagonist oral anticoagulants (NOACs) are alternatives to warfarin in patients with nonvalvular atrial fibrillation. Randomized trials compared NOACs with warfarin, but none have compared individual NOACs against each other for safety and effectiveness. METHODS: We performed a retrospective new-user cohort study of patients with nonvalvular atrial fibrillation enrolled in US Medicare who initiated warfarin (n = 183,318), or a standard dose of dabigatran (150 mg twice daily; n = 86,198), rivaroxaban (20 mg once daily; n = 106,389), or apixaban (5 mg twice daily; n = 73,039) between October 2010 and September 2015. Propensity score-adjusted Cox proportional hazards regression was used to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for the outcomes of thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality, comparing each NOAC with warfarin, and with each other NOAC. RESULTS: Compared with warfarin, each NOAC was associated with reduced risks of thromboembolic stroke (20%-29% reduction; P = .002 [dabigatran], P < 0.001 [rivaroxaban, apixaban]), intracranial hemorrhage (35%-62% reduction; P < 0.001 [each NOAC]), and mortality (19%-34% reduction; P < .001 [each NOAC]). The NOACs were similar for thromboembolic stroke but rivaroxaban was associated with increased risks of intracranial hemorrhage (vs dabigatran: HR = 1.71; 95% CI, 1.35-2.17), major extracranial bleeding (vs dabigatran: HR = 1.32; 95% CI, 1.21-1.45; vs apixaban: HR = 2.70; 95% CI, 2.38-3.05), and death (vs dabigatran: HR = 1.12; 95% CI, 1.01-1.24; vs apixaban: HR = 1.23; 95% CI, 1.09-1.38). Dabigatran was associated with reduced risk of intracranial hemorrhage (HR = 0.70; 95% CI ,0.53-0.94) and increased risk of major extracranial bleeding (HR = 2.04; 95% CI, 1.78-2.32) compared with apixaban. CONCLUSIONS: Among patients treated with standard-dose NOAC for nonvalvular atrial fibrillation and warfarin users with similar baseline characteristics, dabigatran, rivaroxaban, and apixaban were associated with a more favorable benefit-harm profile than warfarin. Among NOAC users, dabigatran and apixaban were associated with a more favorable benefit-harm profile than rivaroxaban.

Data Mining for Adverse Drug Events With a Propensity Score-matched Tree-based Scan Statistic.

The tree-based scan statistic is a statistical data mining tool that has been used for signal detection with a self-controlled design in vaccine safety studies. This disproportionality statistic adjusts for multiple testing in evaluation of thousands of potential adverse events. However, many drug safety questions are not well suited for self-controlled analysis. We propose a method that combines tree-based scan statistics with propensity score-matched analysis of new initiator cohorts, a robust design for investigations of drug safety. We conducted plasmode simulations to evaluate performance. In multiple realistic scenarios, tree-based scan statistics in cohorts that were propensity score matched to adjust for confounding outperformed tree-based scan statistics in unmatched cohorts. In scenarios where confounding moved point estimates away from the null, adjusted analyses recovered the prespecified type 1 error while unadjusted analyses inflated type 1 error. In scenarios where confounding moved point estimates toward the null, adjusted analyses preserved power, whereas unadjusted analyses greatly reduced power. Although complete adjustment of true confounders had the best performance, matching on a moderately mis-specified propensity score substantially improved type 1 error and power compared with no adjustment. When there was true elevation in risk of an adverse event, there were often co-occurring signals for clinically related concepts. TreeScan with propensity score matching shows promise as a method for screening and prioritization of potential adverse events. It should be followed by clinical review and safety studies specifically designed to quantify the magnitude of effect, with confounding control targeted to the outcome of interest.