Tree-based scan statistics to generate drug repurposing hypotheses: a test case using sodium-glucose cotransporter-2 inhibitors.

Most drug repurposing studies using real-world data focused on validating, instead of generating, hypotheses. We used tree-based scan statistics to generate repurposing hypotheses for sodium-glucose cotransporter-2 inhibitors (SGLT2i). We used an active-comparator, new-user design to create a 1:1 propensity-score matched cohort of SGLT2i and dipeptidyl peptidase-4 inhibitors (DPP4i) initiators in the MerativeTM MarketScan® Research Databases. Tree-based scan statistics were estimated across an ICD-10-CM-based hierarchical outcome tree using incident outcomes identified from hospital and outpatient diagnoses. We used an adjusted P≤0.01 as the threshold for statistical alert to prioritize associations for evaluation as repurposing signals. We varied the analyses by tree size, scanning level, and clinical settings for outcomes. There were 80,510 matched SGLT2i-DPP4i initiator pairs with 215,333 outcomes among SGLT2i initiators and 223,428 outcomes among DPP4i initiators. There were 18 prioritized associations, which included chronic kidney disease (P=0.0001), an expected signal, and anemia (P=0.0001). Heart failure (P=0.0167), another expected signal, was identified slightly beyond the statistical alert threshold. Narrowing the outcome tree, scanning at different tree levels, and including outcomes from different clinical settings influenced the scan statistics. We identified signals aligning with recently approved indications of SGLT2i, plus potential repurposing signals supported by existing evidence but requiring future validation.

Impact of the Breast Cancer Index for Extended Endocrine Decision-Making: First Results of the Prospective BCI Registry Study.

BACKGROUND: The Breast Cancer Index (BCI) test assay provides an individualized risk of late distant recurrence (5-10 years) and predicts the likelihood of benefitting from extended endocrine therapy (EET) in hormone receptor-positive early-stage breast cancer. This analysis aimed to assess the impact of BCI on EET decision-making in current clinical practice. METHODS: The BCI Registry study evaluates long-term outcomes, decision impact, and medication adherence in patients receiving BCI testing as part of routine clinical care. Physicians and patients completed pre-BCI and post-BCI test questionnaires to assess a range of questions, including physician decision-making and confidence regarding EET; patient preferences and concerns about the cost, side effects, drug safety, and benefit of EET; and patient satisfaction regarding treatment recommendations. Pre-BCI and post-BCI test responses were compared using McNemar's test and Wilcoxon signed rank test. RESULTS: Pre-BCI and post-BCI questionnaires were completed for 843 physicians and 823 patients. The mean age at enrollment was 65 years, and 88.4% of patients were postmenopausal. Of the tumors, 74.7% were T1, 53.4% were grade 2, 76.0% were N0, and 13.8% were HER2-positive. Following BCI testing, physicians changed EET recommendations in 40.1% of patients (P<.0001), and 45.1% of patients changed their preferences for EET (P<.0001). In addition, 38.8% of physicians felt more confident in their recommendation (P<.0001), and 41.4% of patients felt more comfortable with their EET decision (P<.0001). Compared with baseline, significantly more patients were less concerned about the cost (20.9%; P<.0001), drug safety (25.4%; P=.0014), and benefit of EET (29.3%; P=.0002). CONCLUSIONS: This analysis in a large patient cohort of the BCI Registry confirms and extends previous findings on the significant decision-making impact of BCI on EET. Incorporating BCI into clinical practice resulted in changes in physician recommendations, increased physician confidence, improved patient satisfaction, and reduced patient concerns regarding the cost, drug safety, and benefit of EET.

Imputing missing covariates in time-to-event analysis within distributed research networks: A simulation study.

PURPOSE: In distributed research network (DRN) settings, multiple imputation cannot be directly implemented because pooling individual-level data are often not feasible. The performance of multiple imputation in combination with meta-analysis is not well understood within DRNs. METHODS: To evaluate the performance of imputation for missing baseline covariate data in combination with meta-analysis for time-to-event analysis within DRNs, we compared two parametric algorithms including one approximated linear imputation model (Approx), and one nonlinear substantive model compatible imputation model (SMC), as well as two non-parametric machine learning algorithms including random forest (RF), and classification and regression trees (CART), through simulation studies motivated by a real-world data set. RESULTS: Under the setting with small effect sizes (i.e., log-Hazard ratios [logHR]) and homogeneous missingness mechanisms across sites, all imputation methods produced unbiased and more efficient estimates while the complete-case analysis could be biased and inefficient; and under heterogeneous missingness mechanisms, estimates with RF method could have higher efficiency. Estimates from the distributed imputation combined by meta-analysis were similar to those from the imputation using pooled data. When logHRs were large, the SMC imputation algorithm generally performed better than others. CONCLUSIONS: These findings suggest the validity and feasibility of imputation within DRNs in the presence of missing covariate data in time-to-event analysis under various settings. The performance of the four imputation algorithms varies with the effect sizes and level of missingness.

Association of Antidepressant Prescription Filling With Treatment Indication and Prior Prescription Filling Behaviors and Medication Experiences.

BACKGROUND: Given the wide range of uses for antidepressants, understanding indication-specific patterns of prescription filling for antidepressants provide valuable insights into how patients use these medications in real-world settings. OBJECTIVE: The objective of this study was to determine the association of antidepressant prescription filling with treatment indication, as well as prior prescription filling behaviors and medication experiences. DESIGN: This retrospective cohort study took place in Quebec, Canada. PARTICIPANTS: Adults with public drug insurance prescribed antidepressants using MOXXI (Medical Office of the XXIst Century)-an electronic prescribing system requiring primary care physicians to document treatment indications and reasons for prescription stops or changes. MEASURES: MOXXI provided information on treatment indications, past prescriptions, and prior medication experiences (treatment ineffectiveness and adverse drug reactions). Linked claims data provided information on dispensed medications and other patient-related factors. Multivariable logistic regression models estimated the independent association of not filling an antidepressant prescription (within 90 d) with treatment indication and patients' prior prescription filling behaviors and medication experiences. RESULTS: Among 38,751 prescriptions, the prevalence of unfilled prescriptions for new and ongoing antidepressant therapy was 34.2% and 4.1%, respectively. Compared with depression, odds of not filling an antidepressant prescription varied from 0.74 to 1.57 by indication and therapy status. The odds of not filling an antidepressant prescription was higher among adults filling < 50% of their medication prescriptions in the past year and adults with an antidepressant prescription stopped or changed in the past year due to treatment ineffectiveness. CONCLUSION: Antidepressant prescription filling behaviors differed by treatment indication and were lower among patients with a history of poor prescription filling or ineffective treatment with antidepressants.

Strategies for lung- and diaphragm-protective ventilation in acute hypoxemic respiratory failure: a physiological trial.

BACKGROUND: Insufficient or excessive respiratory effort during acute hypoxemic respiratory failure (AHRF) increases the risk of lung and diaphragm injury. We sought to establish whether respiratory effort can be optimized to achieve lung- and diaphragm-protective (LDP) targets (esophageal pressure swing - 3 to - 8 cm H2O; dynamic transpulmonary driving pressure ≤ 15 cm H2O) during AHRF. METHODS: In patients with early AHRF, spontaneous breathing was initiated as soon as passive ventilation was not deemed mandatory. Inspiratory pressure, sedation, positive end-expiratory pressure (PEEP), and sweep gas flow (in patients receiving veno-venous extracorporeal membrane oxygenation (VV-ECMO)) were systematically titrated to achieve LDP targets. Additionally, partial neuromuscular blockade (pNMBA) was administered in patients with refractory excessive respiratory effort. RESULTS: Of 30 patients enrolled, most had severe AHRF; 16 required VV-ECMO. Respiratory effort was absent in all at enrolment. After initiating spontaneous breathing, most exhibited high respiratory effort and only 6/30 met LDP targets. After titrating ventilation, sedation, and sweep gas flow, LDP targets were achieved in 20/30. LDP targets were more likely to be achieved in patients on VV-ECMO (median OR 10, 95% CrI 2, 81) and at the PEEP level associated with improved dynamic compliance (median OR 33, 95% CrI 5, 898). Administration of pNMBA to patients with refractory excessive effort was well-tolerated and effectively achieved LDP targets. CONCLUSION: Respiratory effort is frequently absent  under deep sedation but becomes excessive when spontaneous breathing is permitted in patients with moderate or severe AHRF. Systematically titrating ventilation and sedation can optimize respiratory effort for lung and diaphragm protection in most patients. VV-ECMO can greatly facilitate the delivery of a LDP strategy. TRIAL REGISTRATION: This trial was registered in Clinicaltrials.gov in August 2018 (NCT03612583).

Diaphragm function in acute respiratory failure and the potential role of phrenic nerve stimulation.

PURPOSE OF REVIEW: The aim of this review was to describe the risk factors for developing diaphragm dysfunction, discuss the monitoring techniques for diaphragm activity and function, and introduce potential strategies to incorporate diaphragm protection into conventional lung-protective mechanical ventilation strategies. RECENT FINDINGS: It is increasingly apparent that an approach that addresses diaphragm-protective ventilations goals is needed to optimize ventilator management and improve patient outcomes. Ventilator-induced diaphragm dysfunction (VIDD) is common and is associated with increased ICU length of stay, prolonged weaning and increased mortality. Over-assistance, under-assistance and patient-ventilator dyssynchrony may have important downstream clinical consequences related to VIDD. Numerous monitoring techniques are available to assess diaphragm function, including respiratory system pressures, oesophageal manometry, diaphragm ultrasound and electromyography. Novel techniques including phrenic nerve stimulation may facilitate the achievement of lung and diaphragm-protective goals for mechanical ventilation. SUMMARY: Diaphragm protection is an important consideration in optimizing ventilator management in patients with acute respiratory failure. The delicate balance between lung and diaphragm-protective goals is challenging. Phrenic nerve stimulation may be uniquely situated to achieve and balance these two commonly conflicting goals.

Can Hyperparameter Tuning Improve the Performance of a Super Learner?: A Case Study.

BACKGROUND: Super learning is an ensemble machine learning approach used increasingly as an alternative to classical prediction techniques. When implementing super learning, however, not tuning the hyperparameters of the algorithms in it may adversely affect the performance of the super learner. METHODS: In this case study, we used data from a Canadian electronic prescribing system to predict when primary care physicians prescribed antidepressants for indications other than depression. The analysis included 73,576 antidepressant prescriptions and 373 candidate predictors. We derived two super learners: one using tuned hyperparameter values for each machine learning algorithm identified through an iterative grid search procedure and the other using the default values. We compared the performance of the tuned super learner to that of the super learner using default values ("untuned") and a carefully constructed logistic regression model from a previous analysis. RESULTS: The tuned super learner had a scaled Brier score (R) of 0.322 (95% [confidence interval] CI = 0.267, 0.362). In comparison, the untuned super learner had a scaled Brier score of 0.309 (95% CI = 0.256, 0.353), corresponding to an efficiency loss of 4% (relative efficiency 0.96; 95% CI = 0.93, 0.99). The previously-derived logistic regression model had a scaled Brier score of 0.307 (95% CI = 0.245, 0.360), corresponding to an efficiency loss of 5% relative to the tuned super learner (relative efficiency 0.95; 95% CI = 0.88, 1.01). CONCLUSIONS: In this case study, hyperparameter tuning produced a super learner that performed slightly better than an untuned super learner. Tuning the hyperparameters of individual algorithms in a super learner may help optimize performance.

A novel non-invasive method to detect excessively high respiratory effort and dynamic transpulmonary driving pressure during mechanical ventilation.

BACKGROUND: Excessive respiratory muscle effort during mechanical ventilation may cause patient self-inflicted lung injury and load-induced diaphragm myotrauma, but there are no non-invasive methods to reliably detect elevated transpulmonary driving pressure and elevated respiratory muscle effort during assisted ventilation. We hypothesized that the swing in airway pressure generated by respiratory muscle effort under assisted ventilation when the airway is briefly occluded (ΔPocc) could be used as a highly feasible non-invasive technique to screen for these conditions. METHODS: Respiratory muscle pressure (Pmus), dynamic transpulmonary driving pressure (ΔPL,dyn, the difference between peak and end-expiratory transpulmonary pressure), and ΔPocc were measured daily in mechanically ventilated patients in two ICUs in Toronto, Canada. A conversion factor to predict ΔPL,dyn and Pmus from ΔPocc was derived and validated using cross-validation. External validity was assessed in an independent cohort (Nanjing, China). RESULTS: Fifty-two daily recordings were collected in 16 patients. In this sample, Pmus and ΔPL were frequently excessively high: Pmus exceeded 10 cm H2O on 84% of study days and ΔPL,dyn exceeded 15 cm H2O on 53% of study days. ΔPocc measurements accurately detected Pmus > 10 cm H2O (AUROC 0.92, 95% CI 0.83-0.97) and ΔPL,dyn > 15 cm H2O (AUROC 0.93, 95% CI 0.86-0.99). In the external validation cohort (n = 12), estimating Pmus and ΔPL,dyn from ΔPocc measurements detected excessively high Pmus and ΔPL,dyn with similar accuracy (AUROC ≥ 0.94). CONCLUSIONS: Measuring ΔPocc enables accurate non-invasive detection of elevated respiratory muscle pressure and transpulmonary driving pressure. Excessive respiratory effort and transpulmonary driving pressure may be frequent in spontaneously breathing ventilated patients.

Assessing the accuracy of using diagnostic codes from administrative data to infer antidepressant treatment indications: a validation study.

PURPOSE: To assess the accuracy of using diagnostic codes from administrative data to infer treatment indications for antidepressants prescribed in primary care. METHODS: Validation study of administrative diagnostic codes for 13 plausible indications for antidepressants compared with physician-documented treatment indications from an indication-based electronic prescribing system in Quebec, Canada. The analysis included all antidepressant prescriptions written by primary care physicians between January 1, 2003 and December 31, 2012 using the electronic prescribing system. Patient prescribed antidepressants were linked to physician claims and hospitalization data to obtain all diagnoses recorded in the past year. RESULTS: Diagnostic codes had poor sensitivity for all treatment indications, ranging from a high of only 31.2% (95% CI, 26.8%-35.9%) for anxiety/stress disorders to as low as 1.3% (95% CI, 0.0%-5.2%) for sexual dysfunction. Sensitivity was notably worse among older patients and patients with more chronic comorbidities. Physician claims data were a better source of diagnostic codes for antidepressant treatment indications than hospitalization data. CONCLUSIONS: Administrative diagnostic codes are poor proxies for antidepressant treatment indications. Future work should determine whether the use of other variables in administrative data besides diagnostic codes can improve the ability to predict antidepressant treatment indications.

A Dynamic Predictive Model for Progression of CKD.

BACKGROUND: Predicting the progression of chronic kidney disease (CKD) is vital for clinical decision making and patient-provider communication. We previously developed an accurate static prediction model that used single-timepoint measurements of demographic and laboratory variables. STUDY DESIGN: Development of a dynamic predictive model using demographic, clinical, and time-dependent laboratory data from a cohort of patients with CKD stages 3 to 5. SETTING & PARTICIPANTS: We studied 3,004 patients seen April 1, 2001, to December 31, 2009, in the outpatient CKD clinic of Sunnybrook Hospital in Toronto, Canada. CANDIDATE PREDICTORS: Age, sex, and urinary albumin-creatinine ratio at baseline. Estimated glomerular filtration rate (eGFR), serum albumin, phosphorus, calcium, and bicarbonate values as time-dependent predictors. OUTCOMES: Treated kidney failure, defined by initiation of dialysis therapy or kidney transplantation. ANALYTICAL APPROACH: We describe a dynamic (latest-available-measurement) prediction model using time-dependent laboratory values as predictors of outcome. Our static model included all 8 candidate predictors. The latest-available-measurement model includes age and the latter 5 variables as time-dependent predictors. We used Cox proportional hazards models for time to kidney failure and compared discrimination, calibration, model fit, and net reclassification for the models. RESULTS: We studied 3,004 patients, who had 344 kidney failure events over a median follow-up of 3 years and an average of 5 clinic visits. eGFR was more strongly associated with kidney failure in the latest-available-measurement model versus the baseline visit static model (HR, 0.44 vs 0.65). The association of calcium level was unchanged, but male sex and phosphorus, albumin, and bicarbonate levels were no longer significant. Discrimination and goodness of fit showed incremental improvement with inclusion of time-dependent covariates (integrated discrimination improvement, 0.73%; 95% CI, 0.56%-0.90%). LIMITATIONS: Our data were derived from a nephrology clinic at a single center. We were unable to include time-dependent changes in albuminuria. CONCLUSIONS: A latest-available-measurement predictive model with eGFR as a time-dependent predictor can incrementally improve risk prediction for kidney failure over a static model with only a single eGFR.