Impact of augmented intelligence (AI) on identification and management of depression in oncology

e14059

Background: Depression is common in patients with cancer and is associated with worse cancer treatment outcomes. Depression is often underdiagnosed/treated as cancer clinicians are focused on the complex aspects of therapy and care coordination. AI has a potential application in the identification of patients at high risk for depression. Jvion has developed a prescriptive analytics solution (the Machine), which uses AI algorithms and machine learning techniques applied to combined clinical and exogenous datasets to identify patients with a propensity for poor clinical outcomes. The Machine was applied to depression risk (within next 6 months), and recommended patient-specific, dynamic, and actionable insights. While the Machine requires no additional documentation within the electronic health record (EHR) to generate its insights, those insights can be integrated back in to any EHR. Herein, we report the results of a pilot study evaluating the impact of AI-driven insights on depression screening and management at a single oncology practice. Methods: All patients were scored weekly using the Machine depression vector. The Machine risk-stratified the patients and generated recommendations for the provider to consider as they developed a care plan. Patients identified as “at risk” by the Machine were assessed for depression (PHQ-9) by the clinical team regardless of prior screening results. The rate per 1000 unique patients per month (PPM) of depression screenings, case management evaluations, and antidepressant prescriptions were calculated for the 5 months prior to and 17 months post deployment of the Machine in the practice. Results: The oncology practice has 21 providers managing an average of 4329 unique PPM. The mean rate of depression screenings increased from 6.0 per 1000 PPM pre- deployment to 16.2 per 1000 PPM post deployment (+271%). The downstream workflow outcomes of case management evaluations increased from 11.6 to 21.4 per 1000 PPM (+184%) and antidepressant prescriptions increased from 9.2 to 15.5 per 1000 PPM (+168%) pre and post-implementation respectively. The providers reported high satisfaction with the use of the AI solution in depression screening. Conclusions: This oncology practice found deployment of the Jvion AI solution to be feasible. The Machine-generated insights for depression risk were actionable, could be incorporated into workflow, and increased the number of patients identified. If confirmed in larger studies, AI-driven insights may improve the identification and management of depression in patients with cancer.

Impact of augmented intelligence (AI) on utilization of palliative care (PC) services in oncology

12015

Background: Timely integration of palliative care in the management of patients with advanced cancer is a quality benchmark in oncology. However, PC is often underutilized as evidenced by delays in identification of appropriate patients, in referrals to a PC service, and in enrollment to hospice. Jvion has developed a prescriptive analytics solution, the Machine, which combines AI algorithms with machine learning techniques and applies them to clinical and exogenous datasets to identify patients with a propensity for poor outcomes. The Machine was applied to risk for patients’ mortality within next 30 days, and recommended patient-specific, dynamic, and actionable insights. Use of the Machine requires no additional documentation within the electronic health record (EHR) and the insights generated can be integrated back in to any EHR to help inform the care plan. Herein, we report the results of a study evaluating the impact of AI-driven insights on PC utilization at a large community oncology practice. Methods: All patients were scored weekly using the Machine PC vector. The Machine risk stratified the patients and generated recommendations for the provider to consider as they developed a care plan. Patients identified as “at risk” by the Machine were assessed for a supportive care visit (PC referral) and then were referred as deemed clinically appropriate. The average monthly rates of PC consults and hospice referrals were calculated 5 months prior to and for 17 months after the launch of the Machine in the practice. Results: The oncology practice has 21 providers managing an average of 4329 unique patients per month (PPM). The mean rate of PC consults increased from 17.3 to 29.1 per 1000 PPM pre and post Machine deployment respectively (+168%). The mean monthly rate of hospice referrals increased by 8-fold from 0.2 to 1.6 per 1000 PPM pre and post deployment respectively. Eliminating the first 6 months of Machine deployment to account for user learning curve, the mean rates of monthly PC consults nearly doubled over baseline to 33.0, and hospice referrals rose 12-fold to 2.4 per 1000 patients in months 7-17 post Machine deployment. Conclusions: This oncology practice found deployment of this novel AI solution to be feasible and effective at generating actionable insights. These AI driven insights could be incorporated into workflow and improved the decision-making for whether and when a patient should be referred to PC and/or hospice services for end of life care. Further study is needed to confirm the value of AI for management of cancer patients at end of life.

Gender-based disparities in clinical trials supporting FDA approval of oncology drugs

2058

Background: Adequate gender representation in clinical trials of new drugs is critical in order to accurately detect possible differences in response and toxicity (Özdemir et al, JCO 2018). The under-representation of women in oncology clinical trials has been previously described, however data on registrational trials, which are the basis for drug approval and inform the prescribing information, is lacking. We conducted an analysis of the trials supporting Food and Drug Administration approval of oncology drugs over a 5-year period to evaluate the representation of women vs. men. Methods: Prescribing information for novel new drugs approved from 2014-2018 was reviewed for the proportions of men and women in the evaluable population of the supporting clinical trials. Sex-specific cancers were excluded. Prevalence estimates for the indications were obtained from the Surveillance, Epidemiology and End Results database and the published literature. A participation to prevalence ratio (PPR) was calculated for each trial by dividing the percentage of women in the trial by the percentage of women in the disease population. A PPR value closer to unity represents even gender distribution and the range 0.8-1.2 is considered to reflect an acceptable representation of women. Data are presented using descriptive statistics. Results: A total of 46 oncology drugs were approved based on 56 trials enrolling 13,862 patients (7941 [57%] men; 5,921 [43%] women). Of the 56 trials, 38 (68%) had a PPR within the 0.8-1.2 range, 15 (27%) fell between 0.4-0.7, and 3 (5%) had a PPR of 1.3. The proportion of trials with unbalanced gender representation was comparable for hematological malignancy and solid tumor indications and did not improve over time. Fewer unbalanced trials were Phase III or employed a randomized design. Nine of the 18 (50%) unbalanced trials enrolled <100 subjects, compared to 3 of the 38 (8%) balanced trials. Conclusions: A third of registrational trials for oncology drugs lacked balanced gender distribution. Of the trials lacking balance, the vast majority (80%) had under-representation of women. Phase I-II trials and smaller trials had greater gender disparity, a concerning finding in a precision medicine environment where an increasing number of registration trials have double digit accrual. Further research is needed to understand the implications of unbalanced gender accrual in registrational trials, and to develop strategies for preventing disparities.

Comparing Inverse Probability of Treatment Weighting and Instrumental Variable Methods for the Evaluation of Adenosine Diphosphate Receptor Inhibitors After Percutaneous Coronary Intervention

IMPORTANCE

There is increasing interest in performing comparative effectiveness analyses in large observational databases, yet these analyses must adjust for treatment selection issues.

OBJECTIVES

To conduct comparative safety and efficacy analyses of prasugrel vs clopidogrel bisulfate after percutaneous coronary intervention and to evaluate inverse probability of treatment weighting (a propensity score method) and instrumental variable methods.

DESIGN, SETTING, AND PARTICIPANTS

This study used data from the Treatment With Adenosine Diphosphate Receptor Inhibitors-Longitudinal Assessment of Treatment Patterns and Events After Acute Coronary Syndrome (TRANSLATE-ACS) study. Included in the study were patients undergoing percutaneous coronary intervention for myocardial infarction, 26.0% of whom received prasugrel. The study dates were April 4, 2010, to October 31, 2012.

EXPOSURES

Choice of initial antiplatelet agent (prasugrel or clopidogrel).

MAIN OUTCOMES AND MEASURES

Safety and efficacy outcomes included 1-year composite major adverse cardiovascular events, moderate to severe bleeding, and stent thrombosis. Hospitalizations for pneumonia, bone fractures, and planned percutaneous coronary intervention were used as the falsification end points.

RESULTS

The study cohort comprised 11 784 participants (mean [SD] age, 60.0 [11.6] years, and 28.0% were female). Using inverse probability of treatment weighting adjustment, prasugrel and clopidogrel had similar major adverse cardiovascular events (hazard ratio [HR], 0.98; 95% CI, 0.83-1.16) and bleeding outcomes (1.18; 0.77-1.80), but prasugrel had a lower rate of stent thrombosis (0.51; 0.31-0.85). Using instrumental variable methods, prasugrel use was associated with a lower rate of the major adverse cardiovascular event end point (HR, 0.68; 95% CI, 0.47-1.00) but nonsignificant differences in the rates of bleeding (0.95; 0.41-2.08) and stent thrombosis (0.67; 0.16-2.00). There was no significant treatment difference noted in any of the falsification end-point rates when analyses were performed using inverse probability of treatment weighting, although the bone fracture end point approached statistical significance. Nevertheless, a lower rate of pneumonia-related hospitalizations was noted in the prasugrel-treated patients when analyses were performed using instrumental variable methods.

CONCLUSIONS AND RELEVANCE

Conclusions regarding the safety and efficacy of antiplatelet therapy varied depending on analytic technique, and none were concordant with the results from randomized trials. In addition, both statistical strategies demonstrated concerning associations when tested in the falsification analyses. A high level of scrutiny and careful attention to assumptions and validity are required when interpreting complex analyses of observational data.

Impact of Proton Pump Inhibitor Use on the Comparative Effectiveness and Safety of Prasugrel Versus Clopidogrel: Insights From the Treatment With Adenosine Diphosphate Receptor Inhibitors: Longitudinal Assessment of Treatment Patterns and Events After Acute Coronary Syndrome (TRANSLATE-ACS) Study

Background

Proton pump inhibitors (PPIs) reduce gastrointestinal bleeding events but may alter clopidogrel metabolism. We sought to understand the comparative effectiveness and safety of prasugrel versus clopidogrel in the context of proton pump inhibitor (PPI) use.

Methods and Results

Using data on 11 955 acute myocardial infarction (MI) patients treated with percutaneous coronary intervention at 233 hospitals and enrolled in the TRANSLATE‐ACS study, we compared whether discharge PPI use altered the association of 1‐year adjusted risks of major adverse cardiovascular events (MACE; death, MI, stroke, or unplanned revascularization) and Global Use of Strategies To Open Occluded Arteries (GUSTO) moderate/severe bleeding between prasugrel‐ and clopidogrel‐treated patients. Overall, 17% of prasugrel‐treated and 19% of clopidogrel‐treated patients received a PPI at hospital discharge. At 1 year, patients discharged on a PPI versus no PPI had higher risks of MACE (adjusted hazard ratio [HR] 1.38, 95% confidence interval [CI] 1.21‐1.58) and GUSTO moderate/severe bleeding (adjusted HR 1.55, 95% CI 1.15‐2.09). Risk of MACE was similar between prasugrel and clopidogrel regardless of PPI use (adjusted HR 0.88, 95% CI 0.62‐1.26 with PPI, adjusted HR 1.07, 95% CI 0.90‐1.28 without PPI, interaction P=0.31). Comparative bleeding risk associated with prasugrel versus clopidogrel use differed based on PPI use but did not reach statistical significance (adjusted HR 0.73, 95% CI 0.36‐1.48 with PPI, adjusted HR 1.34, 95% CI 0.79‐2.27 without PPI, interaction P=0.17).

Conclusions

PPIs did not significantly affect the MACE and bleeding risk associated with prasugrel use, relative to clopidogrel.

Clinical Trial Registration

URL: https://www.clinicaltrials.gov. Unique identifier: NCT01088503.

Contemporary use of platelet function and pharmacogenomic testing among patients with acute myocardial infarction undergoing percutaneous coronary intervention in the United States

BACKGROUND

Although platelet function and pharmacogenomic testing have been studied in clinical trials, their adoption into contemporary practice is unknown.

METHODS

We studied patterns of platelet function and pharmacogenomic testing among 10,048 patients with acute myocardial infarction treated with percutaneous coronary intervention at 226 US hospitals in the TRANSLATE-ACS observational study between April 2010 and October 2012, excluding those receiving research protocol-mandated testing. Inverse probability-weighted propensity adjustment was used to compare 1-year bleeding and major adverse cardiac event risks between patients with and without testing.

RESULTS

Overall, 337 (3.4%) patients underwent predischarge platelet function testing, whereas 85 (0.9%) underwent pharmacogenomic testing; 82% and 93% of hospitals never performed any platelet function or pharmacogenomic testing, respectively. Patients undergoing testing were more likely to be on an adenosine diphosphate receptor inhibitor preadmission or to have percutaneous coronary intervention of a previously treated lesion. Tested patients were more likely than nontested patients to be switched from clopidogrel to prasugrel/ticagrelor (25.7% vs 9.7%, P < .001) and were more likely to be on prasugrel/ticagrelor 6 months postdischarge (33.8% vs 25.1%, P < .001). No significant differences in 1-year bleeding and major adverse cardiac event risks were observed between tested and nontested patients (adjusted hazard ratios 1.06 [95% CI 0.68-1.65] and 1.21 [95% CI 0.94-1.54], respectively).

CONCLUSIONS

Platelet function and pharmacogenomic testing are rarely performed in contemporary myocardial infarction patients in the United States. When tested, patients were more likely to be treated with higher-potency adenosine diphosphate receptor inhibitors, yet no significant differences in longitudinal outcomes were observed.

Early Medication Nonadherence After Acute Myocardial Infarction: Insights into Actionable Opportunities From the TReatment with ADP receptor iNhibitorS: Longitudinal Assessment of Treatment Patterns and Events after Acute Coronary Syndrome (TRANSLATE-ACS) Study

BACKGROUND

Nonadherence to prescribed evidence-based medications after acute myocardial infarction (MI) can contribute to worse outcomes and higher costs. We sought to better understand the modifiable factors contributing to early nonadherence of evidence-based medications after acute MI.

METHODS AND RESULTS

We assessed 7425 acute MI patients treated with percutaneous coronary intervention at 216 US hospitals participating in TReatment with ADP receptor iNhibitorS: Longitudinal Assessment of Treatment Patterns and Events after Acute Coronary Syndrome (TRANSLATE-ACS) between April 2010 and May 2012. Using the validated Morisky instrument to assess cardiovascular medication adherence at 6 weeks post MI, we stratified patients into self-reported high (score, 8), moderate (score, 6-7), and low (score, <6) adherence groups. Moderate and low adherence was reported in 25% and 4% of patients, respectively. One third of low adherence patients described missing doses of antiplatelet therapy at least twice a week after percutaneous coronary intervention. Signs of depression and patient-reported financial hardship because of medication expenses were independently associated with a higher likelihood of medication nonadherence. Patients were more likely to be adherent at 6 weeks if they had follow-up appointments made before discharge and had a provider explain potential side effects of their medications. Lower medication adherence may be associated with a higher risk of 3-month death/readmission (adjusted hazard ratio, 1.35; 95% confidence interval, 0.98-1.87) although this did not reach statistical significance.

CONCLUSIONS

Even early after MI, a substantial proportion of patients report suboptimal adherence to prescribed medications. Tailored patient education and pre discharge planning may represent actionable opportunities to optimize patient adherence and clinical outcomes.

CLINICAL TRIAL REGISTRATION

URL: http://www.clinicaltrials.gov. Unique identifier: NCT01088503.

Persistence with secondary prevention medications after acute myocardial infarction: Insights from the TRANSLATE-ACS study

BACKGROUND

Persistent use of secondary prevention therapies after acute myocardial infarction (MI) is critical to optimizing long-term outcomes.

METHODS

Medication persistence was assessed among 7,955 MI patients in 216 hospitals participating in the Treatment with Adenosine Diphosphate Receptor Inhibitors: Longitudinal Assessment of Treatment Patterns and Events after Acute Coronary Syndrome study from 2010 to 2012. Persistence was defined as continuation of aspirin, adenosine diphosphate receptor inhibitors, β-blockers, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers, and statins from discharge to 6 months post-MI. Multivariable logistic regression modeling was used to determine factors associated with nonpersistence, defined as <80% persistence with all medication classes.

RESULTS

Overall, 31% of MI patients stopped taking a least 1 medication by 6 months. The most common reasons cited for medications discontinuation were side effects and physician instruction (57%), whereas financial concerns were cited in 8% overall. After multivariable modeling, black race (odds ratio 1.36, 95% CI 1.15-1.62), older age (odds ratio 1.07, 95% CI 1.02-1.12), atrial fibrillation (odds ratio 1.67, 95% CI 1.33-2.09), dialysis (odds ratio 1.79, 95% CI 1.15-2.78), and depression (odds ratio 1.22, 95% CI 1.02-1.45) were associated with lower likelihood of persistence. Private insurance (odds ratio 0.85, 95% 0.76-0.95), prescription cost assistance (odds ratio 0.63, 95% CI 0.54-0.75), and outpatient follow-up arranged before discharge (odds ratio 0.89, 95% CI 0.80-0.99) were associated with higher persistence.

CONCLUSIONS

Nearly one-third of MI patients are no longer persistent with their prescribed medications by 6 months. Patients at high risk for nonpersistence may be identified by clinical and sociodemographic features. These observations underscore key opportunities to optimize longitudinal use of secondary prevention therapies.

Pharmacodynamic evaluation of switching from ticagrelor to prasugrel in patients with stable coronary artery disease: Results of the SWAP-2 Study (Switching Anti Platelet-2)

OBJECTIVES

The goal of this study was to evaluate the pharmacodynamic effects of switching patients from ticagrelor to prasugrel.

BACKGROUND

Clinicians may need to switch between more potent P2Y12 inhibitors because of adverse effects or switch to the use of a once-daily dosing regimen due to compliance issues.

METHODS

After a 3- to 5-day run-in phase with a ticagrelor 180-mg loading dose (LD) followed by a ticagrelor 90-mg twice-daily maintenance dose (MD), aspirin-treated patients (N = 110) with stable coronary artery disease were randomized to continue ticagrelor or switch to prasugrel 10-mg once-daily MD, with or without a 60-mg LD. Pharmacodynamic assessments were defined according to P2Y12 reaction unit (PRU) (P2Y12 assay) and platelet reactivity index (vasodilator-stimulated phosphoprotein phosphorylation assay) at baseline (before and after the run-in phase) and 2, 4, 24, and 48 h and 7 days after randomization.

RESULTS

Platelet reactivity was significantly greater at 24 and 48 h after switching to prasugrel versus continued therapy with ticagrelor, although to a lesser extent in those receiving an LD. Mean PRU remained significantly higher in the combined prasugrel groups versus the ticagrelor group (least-squares mean difference: 46 [95% confidence interval 25 to 67]) and did not meet the primary noninferiority endpoint (upper limit of the confidence interval ≤45), although PRU in the prasugrel cohort was lower at 7 days than at 24 or 48 h. Accordingly, rates of high on-treatment platelet reactivity were higher at 24 and 48 h in both prasugrel groups. At 7 days, there was no difference in high on-treatment platelet reactivity rate between the combined prasugrel and ticagrelor groups.

CONCLUSIONS

Compared with continued ticagrelor therapy, switching from ticagrelor to prasugrel therapy was associated with an increase in platelet reactivity that was partially mitigated by the administration of an LD.

Effects of pyridoxal-5′-phosphate (MC-1) in patients undergoing high-risk coronary artery bypass surgery: Results of the MEND-CABG randomized study

OBJECTIVE

Coronary artery bypass graft surgery remains associated with significant postoperative cardiovascular morbidity and mortality in high-risk patients. MC-1 (pyridoxal-5'-phosphate monohydrate) inhibits purinergic receptors and intracellular influx of Ca2+, thereby reducing cellular injury during experimental ischemia and reperfusion. The MEND-CABG trial tested the hypothesis that MC-1 reduces cardiovascular morbidity and mortality after coronary artery bypass graft.

METHODS

In a phase 2, double-blinded, placebo-controlled study, 901 patients scheduled for coronary artery bypass graft surgery with cardiopulmonary bypass and at high risk for subsequent cardiac or neurologic complications were randomly assigned to receive oral MC-1 (250 mg or 750 mg/d once daily) or placebo beginning 3 to 10 hours prior to surgery and continued to postoperative day 30.

RESULTS

At 30 days, MC-1 250 mg (compared with placebo) reduced the composite of death, nonfatal cerebral infarction, and nonfatal myocardial infarction by 14.0% (P = .3124) with peak creatinine kinase-myocardial band > or =50 ng/mL (prespecified primary end point); 32.3% (P = .0349) with peak creatinine kinase-myocardial band > or =70 ng/mL; and 37.2% (P = .0283) with peak creatinine kinase-myocardial band > or =100 ng/mL. Myocardial infarctions with peak creatinine kinase-myocardial band> or =100 ng/mL were reduced by 47.2% in the MC-1 250-mg group versus placebo (P = .0083). Greater efficacy was demonstrated with 250 mg than with the 750-mg dose of MC-1.

CONCLUSIONS

In high-risk patients undergoing coronary artery bypass graft, treatment with MC-1 did not significantly affect the prespecified primary end point but was associated with a significant reduction in perioperative myocardial infarction with creatinine kinase-myocardial band > or =100 ng/mL. A larger, well-powered trial is needed to evaluate the cardioprotective effects of MC-1.