Unsupervised clustering of longitudinal clinical measurements in electronic health records.

Longitudinal electronic health records (EHR) can be utilized to identify patterns of disease development and progression in real-world settings. Unsupervised temporal matching algorithms are being repurposed to EHR from signal processing- and protein-sequence alignment tasks where they have shown immense promise for gaining insight into disease. The robustness of these algorithms for classifying EHR clinical data remains to be determined. Timeseries compiled from clinical measurements, such as blood pressure, have far more irregularity in sampling and missingness than the data for which these algorithms were developed, necessitating a systematic evaluation of these methods. We applied 30 state-of-the-art unsupervised machine learning algorithms to 6,912 systematically generated simulated clinical datasets across five parameters. These algorithms included eight temporal matching algorithms with fourteen partitional and eight fuzzy clustering methods. Nemenyi tests were used to determine differences in accuracy using the Adjusted Rand Index (ARI). Dynamic time warping and its lower-bound variants had the highest accuracies across all cohorts (median ARI>0.70). All 30 methods were better at discriminating classes with differences in magnitude compared to differences in trajectory shapes. Missingness impacted accuracies only when classes were different by trajectory shape. The method with the highest ARI was then used to cluster a large pediatric metabolic syndrome (MetS) cohort (N = 43,426). We identified three unique childhood BMI patterns with high average cluster consensus (>70%). The algorithm identified a cluster with consistently high BMI which had the greatest risk of MetS, consistent with prior literature (OR = 4.87, 95% CI: 3.93-6.12). While these algorithms have been shown to have similar accuracies for regular timeseries, their accuracies in clinical applications vary substantially in discriminating differences in shape and especially with moderate to high missingness (>10%). This systematic assessment also shows that the most robust algorithms tested here can derive meaningful insights from longitudinal clinical data.

Using the Electronic Health Record to Develop a Gastric Cancer Risk Prediction Model.

Background and Aims: Gastric cancer (GC) is a leading cause of cancer incidence and mortality globally. Population screening is limited by the low incidence and prevalence of GC in the United States. A risk prediction algorithm to identify high-risk patients allows for targeted GC screening. We aimed to determine the feasibility and performance of a logistic regression model based on electronic health records to identify individuals at high risk for noncardia gastric cancer (NCGC). Methods: We included 614 patients who had a diagnosis of NCGC between ages 40 and 80 years and who were seen at our large tertiary medical center in multiple states between 2010 and 2021. Controls without a diagnosis of NCGC were randomly selected in a 1:10 ratio of cases to controls. Multiple imputation by chained equations for missing data followed by logistic regression on imputed datasets was used to estimate the probability of NCGC. Area under the curve and the 0.632 estimator was used as the estimate for discrimination. Results: The 0.632 estimator value was 0.731, indicating robust model performance. Probability of NCGC was higher with increasing age (odds ratio [OR] = 1.16, 95% confidence interval [CI]: 1.04-1.3), male sex (OR = 1.97; 95% CI: 1.64-2.36), Black (OR = 3.07; 95% CI: 2.46-3.83) or Asian race (OR = 4.39; 95% CI: 2.60-7.42), tobacco use (OR = 1.61; 95% CI: 1.34-1.94), anemia (OR = 1.35; 95% CI: 1.09-1.68), and pernicious anemia (OR = 6.12, 95% CI: 3.42-10.95). Conclusion: We demonstrate the feasibility and good performance of an electronic health record-based logistic regression model for estimating the probability of NCGC. Future studies will refine and validate this model, ultimately identifying a high-risk cohort who could be eligible for NCGC screening.

Iris freckle: a distinct entity.

PURPOSE: To report distinctive clinical and imaging features of iris freckles to differentiate them from iris nevi. DESIGN: Retrospective observational study. SUBJECTS: 53 patients (277 freckles) with incidental iris freckles and 102 patients (104 nevi) with iris nevi that are either clinically stable or pathologically confirmed. METHODS: Patient data were collected from the Department of Ophthalmic Oncology, Cleveland Clinic, Cole Eye Institute database (2012-2023). Lesion characteristics were recorded from slit-lamp examination descriptions and review of colour photographs. Ancillary imaging features observed using anterior segment optical coherence tomography (AS-OCT) and ultrasound biomicroscopy (UBM) were assessed in patients (where available). MAIN OUTCOME MEASURES: Comparison of clinical and imaging features of iris freckles and iris nevi. RESULTS: A total of 277 iris freckles and 104 iris nevi were analysed. Iris freckles were more frequently bilateral (17%; nevi 0%) and multiple (69%; nevi 2%) and located centrally (89%; nevi 17%) compared with iris nevi (p<0.001). The median freckle largest basal diameter and thickness were 0.8 mm (nevi; 2.1 mm, p<0.001) and 0.04 mm (nevi 1.0 mm, p<0.001), respectively. All iris freckles had irregular margins without any secondary effects compared with iris nevi. Iris freckles appeared flat without effacement of iris folds compared with iris nevi on AS-OCT (p<0.001). Iris freckles were not detectable by UBM. Heat map revealed that freckles demonstrated several features with uniform or near uniform values, whereas nevi demonstrated more variability in values across features. CONCLUSIONS: Iris freckles exhibit specific clinical and imaging features reflective of their characteristic histological composition that support their classification as a distinct entity within the spectrum of iris pigmented lesions.

Predictive model for iris melanoma.

AimTo develop a predictive model for the diagnosis of iris melanoma. METHODS: Retrospective consecutive case series that included 100 cases of pathologically confirmed iris melanoma and 112 cases of Iris naevus, either pathological confirmation or documented stability of >1 year. Patient demographic data, features of clinical presentation, tumour characteristics and follow-up were collected. Iris melanoma with ciliary body extension was excluded. Lasso logistic regression with 10-fold cross-validation was used to select the tuning parameter. Discrimination was assessed with the area under the curve (AUC) and calibration by a plot. RESULTS: There was a significant asymmetry in the location of both nevi and melanoma with preference for inferior iris quadrants (83, 74%) and (79, 79%), respectively (p=0.50). Tumour seeding, glaucoma and hyphaema were present only in melanoma. The features that favoured the diagnosis of melanoma were size (increased height (OR 3.35); increased the largest basal diameter (OR 1.64)), pupillary distortion (ectropion uvea or corectopia (OR 2.55)), peripheral extension (angle or iris root involvement (OR 2.83)), secondary effects (pigment dispersion (OR 1.12)) and vascularity (OR 6.79). The optimism-corrected AUC was 0.865. The calibration plot indicated good calibration with most of the points falling near the identity line and the confidence band containing the identity line through most of the range of probabilities. CONCLUSIONS: The predictive model provides direct diagnostic prediction of the lesion being iris melanoma expressed as probability (%). Use of a prediction calculator (app) can enhance decision-making and patient counselling. Further refinements can be undertaken with additional datasets, forming the basis for automated diagnosis.

Surveillance for Metastasis in High-Risk Uveal Melanoma Patients: Standard versus Enhanced Protocols.

PURPOSE: to evaluate the effectiveness of enhanced surveillance protocols (EP) utilizing high frequency (HF) or enhanced modality (EM) compared to the standard protocol (SP) in detecting metastasis and determining their impact on overall survival (OS) in high-risk uveal melanoma (UM) patients. METHODS: A total of 87 consecutive patients with Class 2 (high risk) primary UM were enrolled, with negative baseline systemic staging. The patients underwent systemic surveillance with either SP (hepatic ultrasonography [US] every 6 months) or EP (either HF [US every 3 months] or EM [incorporation hepatic computed tomography/magnetic resonance imaging]) following informed discussion. The main outcome measures were largest diameter of largest hepatic metastasis (LDLM), number of hepatic metastatic lesions, time to detection of metastasis (TDM), and OS. RESULTS: This study revealed significant differences in LDLM between surveillance protocols, with the use of EP detecting smaller metastatic lesions (HF, EM, and SP were 1.5 cm, 1.6 cm, and 6.1 cm, respectively). Patients on the EM protocol had a lower 24-month cumulative incidence of >3 cm metastasis (3.5% EM vs. 39% SP; p = 0.021), while those on the HF protocol had a higher 24-month cumulative incidence of ≤3 cm metastasis compared to SP (31% HF vs. 10% SP; p = 0.017). Hazard of death following metastasis was significantly reduced in the EP (HR: 0.25; 95% CI: 0.07, 0.84), HF (HR: 0.23; 95% CI: 0.06, 0.84), and EM (HR: 0.11; 95% CI: 0.02, 0.5) groups compared to SP. However, TDM and OS did not significantly differ between protocols. CONCLUSIONS: Enhanced surveillance protocols improved early detection of hepatic metastasis in UM patients but did not translate into a survival advantage in our study cohort. However, early detection of metastasis in patients receiving liver-directed therapies may lead to improved overall survival.

Guidelines for Reporting Observational Research in Urology: The Importance of Clear Reference to Causality.

Observational studies often dance around the issue of causality. We propose guidelines to ensure that papers refer to whether or not the study aim is to investigate causality, and suggest language to use and language to avoid.

Piperacillin-Tazobactam Compared With Cefoxitin as Antimicrobial Prophylaxis for Pancreatoduodenectomy: A Randomized Clinical Trial.

Importance: Despite improvements in perioperative mortality, the incidence of postoperative surgical site infection (SSI) remains high after pancreatoduodenectomy. The effect of broad-spectrum antimicrobial surgical prophylaxis in reducing SSI is poorly understood. Objective: To define the effect of broad-spectrum perioperative antimicrobial prophylaxis on postoperative SSI incidence compared with standard care antibiotics. Design, Setting, and Participants: Pragmatic, open-label, multicenter, randomized phase 3 clinical trial at 26 hospitals across the US and Canada. Participants were enrolled between November 2017 and August 2021, with follow-up through December 2021. Adults undergoing open pancreatoduodenectomy for any indication were eligible. Individuals were excluded if they had allergies to study medications, active infections, chronic steroid use, significant kidney dysfunction, or were pregnant or breastfeeding. Participants were block randomized in a 1:1 ratio and stratified by the presence of a preoperative biliary stent. Participants, investigators, and statisticians analyzing trial data were unblinded to treatment assignment. Intervention: The intervention group received piperacillin-tazobactam (3.375 or 4 g intravenously) as perioperative antimicrobial prophylaxis, while the control group received cefoxitin (2 g intravenously; standard care). Main Outcomes and Measures: The primary outcome was development of postoperative SSI within 30 days. Secondary end points included 30-day mortality, development of clinically relevant postoperative pancreatic fistula, and sepsis. All data were collected as part of the American College of Surgeons National Surgical Quality Improvement Program. Results: The trial was terminated at an interim analysis on the basis of a predefined stopping rule. Of 778 participants (378 in the piperacillin-tazobactam group [median age, 66.8 y; 233 {61.6%} men] and 400 in the cefoxitin group [median age, 68.0 y; 223 {55.8%} men]), the percentage with SSI at 30 days was lower in the perioperative piperacillin-tazobactam vs cefoxitin group (19.8% vs 32.8%; absolute difference, -13.0% [95% CI, -19.1% to -6.9%]; P < .001). Participants treated with piperacillin-tazobactam, vs cefoxitin, had lower rates of postoperative sepsis (4.2% vs 7.5%; difference, -3.3% [95% CI, -6.6% to 0.0%]; P = .02) and clinically relevant postoperative pancreatic fistula (12.7% vs 19.0%; difference, -6.3% [95% CI, -11.4% to -1.2%]; P = .03). Mortality rates at 30 days were 1.3% (5/378) among participants treated with piperacillin-tazobactam and 2.5% (10/400) among those receiving cefoxitin (difference, -1.2% [95% CI, -3.1% to 0.7%]; P = .32). Conclusions and Relevance: In participants undergoing open pancreatoduodenectomy, use of piperacillin-tazobactam as perioperative prophylaxis reduced postoperative SSI, pancreatic fistula, and multiple downstream sequelae of SSI. The findings support the use of piperacillin-tazobactam as standard care for open pancreatoduodenectomy. Trial Registration: ClinicalTrials.gov Identifier: NCT03269994.

Optimal predictive probability designs for randomized biomarker-guided oncology trials.

Introduction: Efforts to develop biomarker-targeted anti-cancer therapies have progressed rapidly in recent years. With efforts to expedite regulatory reviews of promising therapies, several targeted cancer therapies have been granted accelerated approval on the basis of evidence acquired in single-arm phase II clinical trials. And yet, in the absence of randomization, patient prognosis for progression-free survival and overall survival may not have been studied under standard of care chemotherapies for emerging biomarker subpopulations prior to the submission of an accelerated approval application. Historical control rates used to design and evaluate emerging targeted therapies often arise as population averages, lacking specificity to the targeted genetic or immunophenotypic profile. Thus, historical trial results are inherently limited for inferring the potential "comparative efficacy" of novel targeted therapies. Consequently, randomization may be unavoidable in this setting. Innovations in design methodology are needed, however, to enable efficient implementation of randomized trials for agents that target biomarker subpopulations. Methods: This article proposes three randomized designs for early phase biomarker-guided oncology clinical trials. Each design utilizes the optimal efficiency predictive probability method to monitor multiple biomarker subpopulations for futility. Only designs with type I error between 0.05 and 0.1 and power of at least 0.8 were considered when selecting an optimal efficiency design from among the candidate designs formed by different combinations of posterior and predictive threshold. A simulation study motivated by the results reported in a recent clinical trial studying atezolizumab treatment in patients with locally advanced or metastatic urothelial carcinoma is used to evaluate the operating characteristics of the various designs. Results: Out of a maximum of 300 total patients, we find that the enrichment design has an average total sample size under the null of 101.0 and a total average sample size under the alternative of 218.0, as compared to 144.8 and 213.8 under the null and alternative, respectively, for the stratified control arm design. The pooled control arm design enrolled a total of 113.2 patients under the null and 159.6 under the alternative, out of a maximum of 200. These average sample sizes that are 23-48% smaller under the alternative and 47-64% smaller under the null, as compared to the realized sample size of 310 patients in the phase II study of atezolizumab. Discussion: Our findings suggest that potentially smaller phase II trials to those used in practice can be designed using randomization and futility stopping to efficiently obtain more information about both the treatment and control groups prior to phase III study.

Cardiovascular risk factors, radiation therapy, and myocardial infarction among lymphoma survivors.

BACKGROUND: Mediastinal radiation is associated with increased risk of myocardial infarction (MI) among non-Hodgkin lymphoma (NHL) survivors. OBJECTIVE: To evaluate how preexisting cardiovascular risk factors (CVRFs) modify the association of mediastinal radiation and MI among a national population of NHL survivors with a range of CVRFs. MATERIAL AND METHODS: Using Danish registries, we identified adults diagnosed with lymphoma 2000-2010. We assessed MI from one year after diagnosis through 2016. We ascertained CVRFs (hypertension, dyslipidemia, and diabetes), vascular disease, and intrinsic heart disease prevalent at lymphoma diagnosis. We used multivariable Cox regression to test the interaction between preexisting CVRFs and receipt of mediastinal radiation on subsequent MI. RESULTS: Among 3151 NHL survivors (median age 63, median follow-up 6.5 years), 96 were diagnosed with MI. Before lymphoma, 32% of survivors had ≥1 CVRF. 8.5% of survivors received mediastinal radiation. In multivariable analysis, we found that mediastinal radiation (HR = 1.96; 95% CI = 1.09-3.52), and presence of ≥1 CVRF (HR = 2.71; 95% CI = 1.77-4.15) were associated with an increased risk of MI. Although there was no interaction on the relative scale (p = 0.14), we saw a clinically relevant absolute increase in risk for patients with CVRF from 10-year of MI of 10.5% without radiation to 29.5% for those undergoing radiation. CONCLUSION: Patients with CVRFs have an importantly higher risk of subsequent MI if they have mediastinal radiation. Routine evaluation of CVRFs and optimal treatment of preexisting cardiovascular disease should continue after receiving cancer therapy. In patients with CVRFs, mediastinal radiation should only be given if oncologic benefit clearly outweighs cardiovascular harm.

Identification of robust deep neural network models of longitudinal clinical measurements.

Deep learning (DL) from electronic health records holds promise for disease prediction, but systematic methods for learning from simulated longitudinal clinical measurements have yet to be reported. We compared nine DL frameworks using simulated body mass index (BMI), glucose, and systolic blood pressure trajectories, independently isolated shape and magnitude changes, and evaluated model performance across various parameters (e.g., irregularity, missingness). Overall, discrimination based on variation in shape was more challenging than magnitude. Time-series forest-convolutional neural networks (TSF-CNN) and Gramian angular field(GAF)-CNN outperformed other approaches (P < 0.05) with overall area-under-the-curve (AUCs) of 0.93 for both models, and 0.92 and 0.89 for variation in magnitude and shape with up to 50% missing data. Furthermore, in a real-world assessment, the TSF-CNN model predicted T2D with AUCs reaching 0.72 using only BMI trajectories. In conclusion, we performed an extensive evaluation of DL approaches and identified robust modeling frameworks for disease prediction based on longitudinal clinical measurements.

A multi-institutional prediction model to estimate the risk of recurrence and mortality after mastectomy for T1-2N1 breast cancer.

BACKGROUND: Post-mastectomy radiation therapy (PMRT) in women with pathologic stage T1-2N1M0 breast cancer is controversial. METHODS: Data from five North American institutions including women undergoing mastectomy without neoadjuvant therapy with pT1-2N1M0 breast cancer treated from 2006 to 2015 were pooled for analysis. Competing-risks regression was performed to identify factors associated with locoregional recurrence (LRR), distant metastasis (DM), overall recurrence (OR), and breast cancer mortality (BCM). RESULTS: A total of 3532 patients were included for analysis with a median follow-up time among survivors of 6.8 years (interquartile range [IQR], 4.5-9.5 years). The 2154 (61%) patients who received PMRT had significantly more adverse risk factors than those patients not receiving PMRT: younger age, larger tumors, more positive lymph nodes, lymphovascular invasion, extracapsular extension, and positive margins (p < .05 for all). On competing risk regression analysis, receipt of PMRT was significantly associated with a decreased risk of LRR (hazard ratio [HR], 0.21; 95% confidence interval [CI], 0.14-0.31; p < .001) and OR (HR, 0.76; 95% CI, 0.62-0.94; p = .011). Model performance metrics for each end point showed good discrimination and calibration. An online prediction model to estimate predicted risks for each outcome based on individual patient and tumor characteristics was created from the model. CONCLUSIONS: In a large multi-institutional cohort of patients, PMRT for T1-2N1 breast cancer was associated with a significant reduction in locoregional and overall recurrence after accounting for known prognostic factors. An online calculator was developed to aid in personalized decision-making regarding PMRT in this population.

Basket Trials: Review of Current Practice and Innovations for Future Trials.

Advances in biology and immunology have elucidated genetic and immunologic origins of cancer. Innovations in sequencing technologies revealed that distinct cancer histologies shared common genetic and immune phenotypic traits. Pharmacologic developments made it possible to target these alterations, yielding novel classes of targeted agents whose therapeutic potential span multiple tumor types. Basket trials, one type of master protocol, emerged as a tool for evaluating biomarker-targeted therapies among multiple tumor histologies. Conventionally conducted within the phase II setting and designed to estimate high and durable objective responses, basket trials pose challenges to statistical design and interpretation of results. This article reviews basket trials implemented in oncology studies and discusses issues related to their statistical design and analysis.

Optimal Sequential Predictive Probability Designs for Early-Phase Oncology Expansion Cohorts.

PURPOSE: The customary approach to early-phase clinical trial design, where the focus is on identification of the maximum tolerated dose, is not always suitable for noncytotoxic or other targeted therapies. Many trials have continued to follow the 3 + 3 dose-escalation design, but with the addition of phase I dose-expansion cohorts to further characterize safety and assess efficacy. Dose-expansion cohorts are not always planned in advance nor rigorously designed. We introduce an approach to the design of phase I expansion cohorts on the basis of sequential predictive probability monitoring. METHODS: Two optimization criteria are proposed that allow investigators to stop for futility to preserve limited resources while maintaining traditional control of type I and type II errors. We demonstrate the use of these designs through simulation, and we elucidate their implementation with a redesign of the phase I expansion cohort for atezolizumab in metastatic urothelial carcinoma. RESULTS: A sequential predictive probability design outperforms Simon's two-stage designs and posterior probability monitoring with respect to both proposed optimization criteria. The Bayesian sequential predictive probability design yields increased power while significantly reducing the average sample size under the null hypothesis in the context of the case study, whereas the original study design yields too low type I error and power. The optimal efficiency design tended to have more desirable properties, subject to constraints on type I error and power, compared with the optimal accuracy design. CONCLUSION: The optimal efficiency design allows investigators to preserve limited financial resources and to maintain ethical standards by halting potentially large dose-expansion cohorts early in the absence of promising efficacy results, while maintaining traditional control of type I and II error rates.

Modeling the Cost-Effectiveness of Adjuvant Osimertinib for Patients with Resected EGFR-mutant Non-Small Cell Lung Cancer.

INTRODUCTION: The epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor osimertinib was recently approved for resected EGFR-mutant stages IB-IIIA non-small cell lung cancer due to improved disease-free survival (DFS) in this population compared with placebo. This study aimed to evaluate the cost-effectiveness (CE) of this strategy. MATERIALS AND METHODS: We constructed a Markov model using post-resection health state transitions with digitized DFS data from the ADAURA trial to compare cost and quality-adjusted life years (QALYs) of 3 years of adjuvant osimertinib versus placebo over a 10-year time horizon. An overall survival (OS) benefit of 5% was assumed. Costs and utility values were derived from Medicare reimbursement data and literature. A CE threshold of 3 times the gross domestic product per capita was used. Sensitivity analyses were performed. RESULTS: The incremental cost-effectiveness ratio for adjuvant osimertinib was $317 119 per QALY-gained versus placebo. Initial costs of osimertinib are higher in years 1-3. Costs due to progressive disease (PD) are higher in the placebo group through the first 6.5 years. Average pre-PD, post-PD, and total costs were $2388, $379 047, and $502 937, respectively, in the placebo group, and $505 775, $255 638, and $800 697, respectively, in the osimertinib group. Sensitivity analysis of OS gains reaches CE with an hazard ratio (HR) of 0.70-0.75 benefit of osimertinib over placebo. A 50% discount to osimertinib drug cost yielded an ICER of $115 419. CONCLUSIONS: Three-years of adjuvant osimertinib is CE if one is willing to pay $317 119 more per QALY-gained. Considerable OS benefit over placebo or other economic interventions will be needed to reach CE.

A Prediction Model to Discriminate Small Choroidal Melanoma from Choroidal Nevus.

Objective: This study aimed to develop a validated machine learning model to diagnose small choroidal melanoma. Design: This is a cohort study. Subjects Participants and/or Controls: The training data included 123 patients diagnosed as small choroidal melanocytic tumor (5.0-16.0 mm in largest basal diameter and 1.0 mm-2.5 mm in height; Collaborative Ocular Melanoma Study criteria). Those diagnosed as melanoma (n = 61) had either documented growth or pathologic confirmation. Sixty-two patients with stable lesions classified as choroidal nevus were used as negative controls. The external validation dataset included 240 patients managed at a different tertiary clinic, also with small choroidal melanocytic tumor, observed for malignant growth. Methods: In the training data, lasso logistic regression was used to select variables for inclusion in the final model for the association with melanoma versus choroidal nevus. Internal and external validation was performed to assess model performance. Main Outcome Measures: The main outcome measure is the predicted probability of small choroidal melanoma. Results: Distance to optic disc ≥3 mm and drusen were associated with decreased odds of melanoma, whereas male versus female sex, increased height, subretinal fluid, and orange pigment were associated with increased odds of choroidal melanoma. The area under the receiver operating characteristic "discrimination value" for this model was 0.880. The top four variables that were most frequently selected for inclusion in the model on internal validation, implying their importance as predictors of melanoma, were subretinal fluid, height, distance to optic disc, and orange pigment. When tested against the validation data, the prediction model could distinguish between choroidal nevus and melanoma with a high discrimination of 0.861. The final prediction model was converted into an online calculator to generate predicted probability of melanoma. Conclusions: To minimize diagnostic uncertainty, a machine learning-based diagnostic prediction calculator can be readily applied for decision-making and counseling patients with small choroidal melanoma.