Multi-dimensional predictors of first drinking initiation and regular drinking onset in adolescence: A prospective longitudinal study.

Early adolescent drinking onset is linked to myriad negative consequences. Using the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA) baseline to year 8 data, this study (1) leveraged best subsets selection and Cox Proportional Hazards regressions to identify the most robust predictors of adolescent first and regular drinking onset, and (2) examined the clinical utility of drinking onset in forecasting later binge drinking and withdrawal effects. Baseline predictors included youth psychodevelopmental characteristics, cognition, brain structure, family, peer, and neighborhood domains. Participants (N=538) were alcohol-naïve at baseline. The strongest predictors of first and regular drinking onset were positive alcohol expectancies (Hazard Ratios [HRs]=1.67-1.87), easy home alcohol access (HRs=1.62-1.67), more parental solicitation (e.g., inquiring about activities; HRs=1.72-1.76), and less parental control and knowledge (HRs=.72-.73). Robust linear regressions showed earlier first and regular drinking onset predicted earlier transition into binge and regular binge drinking (ßs=0.57-0.95). Zero-inflated Poisson regressions revealed that delayed first and regular drinking increased the likelihood (Incidence Rate Ratios [IRR]=1.62 and IRR=1.29, respectively) of never experiencing withdrawal. Findings identified behavioral and environmental factors predicting temporal paths to youthful drinking, dissociated first from regular drinking initiation, and revealed adverse sequelae of younger drinking initiation, supporting efforts to delay drinking onset.

predicTTE: An accessible and optimal tool for time-to-event prediction in neurological diseases.

Time-to-event prediction is a key task for biological discovery, experimental medicine, and clinical care. This is particularly true for neurological diseases where development of reliable biomarkers is often limited by difficulty visualising and sampling relevant cell and molecular pathobiology. To date, much work has relied on Cox regression because of ease-of-use, despite evidence that this model includes incorrect assumptions. We have implemented a set of deep learning and spline models for time-to-event modelling within a fully customizable 'app' and accompanying online portal, both of which can be used for any time-to-event analysis in any disease by a non-expert user. Our online portal includes capacity for end-users including patients, Neurology clinicians, and researchers, to access and perform predictions using a trained model, and to contribute new data for model improvement, all within a data-secure environment. We demonstrate a pipeline for use of our app with three use-cases including imputation of missing data, hyperparameter tuning, model training and independent validation. We show that predictions are optimal for use in downstream applications such as genetic discovery, biomarker interpretation, and personalised choice of medication. We demonstrate the efficiency of an ensemble configuration, including focused training of a deep learning model. We have optimised a pipeline for imputation of missing data in combination with time-to-event prediction models. Overall, we provide a powerful and accessible tool to develop, access and share time-to-event prediction models; all software and tutorials are available at www.predictte.org.

A time-to-event modelling of sputum conversion within two months after antituberculosis initiation among drug-susceptible smear positive pulmonary tuberculosis patients: Implementation of internal and external validation.

Delayed sputum conversion has been associated with a higher risk of treatment failure or relapse among drug susceptible smear-positive pulmonary tuberculosis patients. Several contributing factors have been identified in many studies, but the results varied across regions and countries. Therefore, the current study aimed to develop a predictive model that explained the factors affecting time to sputum conversion within two months after initiating antituberculosis agents among Malaysian with drug-susceptible smear-positive pulmonary tuberculosis patients. Retrospective data of pulmonary tuberculosis patients followed up at a tertiary hospital in the Northern region of Malaysia from 2013 until 2018 were collected and analysed. Nonlinear mixed-effect modelling software (NONMEM 7.3.0) was used to develop parametric survival models. The final model was further validated using Kaplan-Meier-visual predictive check (KM-VPC) approach, kernel-based hazard rate estimation method and sampling-importance resampling (SIR) method. A total of 224 patients were included in the study, with 34.4 % (77/224) of the patients remained positive at the end of 2 months of the intensive phase. Gompertz hazard function best described the data. The hazard of sputum conversion decreased by 39 % and 33 % for moderate and advanced lesions as compared to minimal baseline of chest X-ray severity, respectively (adjusted hazard ratio (aHR), 0.61; 95 % confidence intervals (95 % CI), (0.44-0.84) and 0.67, 95 % CI (0.53-0.84)). Meanwhile, the hazard also decreased by 59 % (aHR, 0.41; 95 % CI, (0.23-0.73)) and 48 % (aHR, 0.52; 95 % CI, (0.35-0.79)) between active and former drug abusers as compared to non-drug abuser, respectively. The successful development of the internally and externally validated final model allows a better estimation of the time to sputum conversion and provides a better understanding of the relationship with its predictors.

Using multistate models with clinical trial data for a deeper understanding of complex disease processes.

A clinical trial represents a large commitment from all individuals involved and a huge financial obligation given its high cost; therefore, it is wise to make the most of all collected data by learning as much as possible. A multistate model is a generalized framework to describe longitudinal events; multistate hazards models can treat multiple intermediate/final clinical endpoints as outcomes and estimate the impact of covariates simultaneously. Proportional hazards models are fitted (one per transition), which can be used to calculate the absolute risks, that is, the probability of being in a state at a given time, the expected number of visits to a state, and the expected amount of time spent in a state. Three publicly available clinical trial datasets, colon, myeloid, and rhDNase, in the survival package in R were used to showcase the utility of multistate hazards models. In the colon dataset, a very well-known and well-used dataset, we found that the levamisole+fluorouracil treatment extended time in the recurrence-free state more than it extended overall survival, which resulted in less time in the recurrence state, an example of the classic "compression of morbidity." In the myeloid dataset, we found that complete response (CR) is durable, patients who received treatment B have longer sojourn time in CR than patients who received treatment A, while the mutation status does not impact the transition rate to CR but is highly influential on the sojourn time in CR. We also found that more patients in treatment A received transplants without CR, and more patients in treatment B received transplants after CR. In addition, the mutation status is highly influential on the CR to transplant transition rate. The observations that we made on these three datasets would not be possible without multistate models. We want to encourage readers to spend more time to look deeper into clinical trial data. It has a lot more to offer than a simple yes/no answer if only we, the statisticians, are willing to look for it.

A competing risks model to estimate the risk of graft failure and patient death after kidney transplantation using continuous donor-recipient age combinations.

Graft failure and recipient death with functioning graft are important competing outcomes after kidney transplantation. Risk prediction models typically censor for the competing outcome thereby overestimating the cumulative incidence. The magnitude of this overestimation is not well-described in real-world transplant data. This retrospective cohort study analyzed data from the European Collaborative Transplant Study (CTS; n = 125 250) and from the American Scientific Registry of Transplant Recipients (SRTR; n = 190 258). Separate cause-specific hazard models, using donor and recipient age as continuous predictors, were developed for graft failure and recipient death. The hazard of graft failure increased quadratically with increasing donor age and decreased decaying with increasing recipient age. The hazard of recipient death increased linearly with increasing donor and recipient age. The cumulative incidence overestimation due to competing risk-censoring was largest in high-risk populations for both outcomes (old donors/recipients), sometimes amounting to 8.4 and 18.8 percentage points for graft failure and recipient death, respectively. In our illustrative model for post-transplant risk prediction, the absolute risk of graft failure and death is overestimated when censoring for the competing event, mainly in older donors and recipients. Prediction models for absolute risks should treat graft failure and death as competing events.

Statistical approaches for component-wise censored composite endpoints.

Composite endpoints defined as the time to the earliest of two or more events are often used as primary endpoints in clinical trials. Component-wise censoring arises when different components of the composite endpoint are censored differently. We focus on a composite of death and a non-fatal event where death time is right censored and the non-fatal event time is interval censored because the event can only be detected during study visits. Such data are most often analysed using methods for right censored data, treating the time the non-fatal event was first detected as the time it occurred. This can lead to bias, particularly when the time between assessments is long. We describe several approaches for estimating the event-free survival curve and the effect of treatment on event-free survival via the hazard ratio that are specifically designed to handle component-wise censoring. We apply the methods to a randomized study of breastfeeding versus formula feeding for infants of mothers infected with human immunodeficiency virus.

Sample Size Calculation Under Nonproportional Hazards Using Average Hazard Ratios.

Many clinical trials assess time-to-event endpoints. To describe the difference between groups in terms of time to event, we often employ hazard ratios. However, the hazard ratio is only informative in the case of proportional hazards (PHs) over time. There exist many other effect measures that do not require PHs. One of them is the average hazard ratio (AHR). Its core idea is to utilize a time-dependent weighting function that accounts for time variation. Though propagated in methodological research papers, the AHR is rarely used in practice. To facilitate its application, we unfold approaches for sample size calculation of an AHR test. We assess the reliability of the sample size calculation by extensive simulation studies covering various survival and censoring distributions with proportional as well as nonproportional hazards (N-PHs). The findings suggest that a simulation-based sample size calculation approach can be useful for designing clinical trials with N-PHs. Using the AHR can result in increased statistical power to detect differences between groups with more efficient sample sizes.

The multivariate Bernoulli detector: change point estimation in discrete survival analysis.

Time-to-event data are often recorded on a discrete scale with multiple, competing risks as potential causes for the event. In this context, application of continuous survival analysis methods with a single risk suffers from biased estimation. Therefore, we propose the multivariate Bernoulli detector for competing risks with discrete times involving a multivariate change point model on the cause-specific baseline hazards. Through the prior on the number of change points and their location, we impose dependence between change points across risks, as well as allowing for data-driven learning of their number. Then, conditionally on these change points, a multivariate Bernoulli prior is used to infer which risks are involved. Focus of posterior inference is cause-specific hazard rates and dependence across risks. Such dependence is often present due to subject-specific changes across time that affect all risks. Full posterior inference is performed through a tailored local-global Markov chain Monte Carlo (MCMC) algorithm, which exploits a data augmentation trick and MCMC updates from nonconjugate Bayesian nonparametric methods. We illustrate our model in simulations and on ICU data, comparing its performance with existing approaches.

A Shared-Frailty Spatial Scan Statistic Model for Time-to-Event Data.

Spatial scan statistics are well-known methods widely used to detect spatial clusters of events. Furthermore, several spatial scan statistics models have been applied to the spatial analysis of time-to-event data. However, these models do not take account of potential correlations between the observations of individuals within the same spatial unit or potential spatial dependence between spatial units. To overcome this problem, we have developed a scan statistic based on a Cox model with shared frailty and that takes account of the spatial dependence between spatial units. In simulation studies, we found that (i) conventional models of spatial scan statistics for time-to-event data fail to maintain the type I error in the presence of a correlation between the observations of individuals within the same spatial unit and (ii) our model performed well in the presence of such correlation and spatial dependence. We have applied our method to epidemiological data and the detection of spatial clusters of mortality in patients with end-stage renal disease in northern France.

Evaluation of the proportional hazards assumption and covariate adjustment methods in comparative surgical observational studies with time-to-event endpoints.

INTRODUCTION: Comparative studies on surgical treatments with time-to-event endpoints have provided substantial evidence for clinical practice, but the accurate use of survival data analysis and the control of confounding bias remain big challenges. METHODS: This was a survey of surgical studies with survival outcomes published in four general medical journals and five general surgical journals in 2021. The two most concerned statistical issues were evaluated, including confounding control by propensity score analysis (PSA) or multivariable analysis and testing of proportional hazards (PH) assumption in Cox model. RESULTS: A total of 74 studies were included, comprising 63 observational studies and 11 randomized controlled trials. Among the observational studies, the proportion of studies utilizing PSA in surgical oncology and non-oncology studies was similar (40.9 % versus 36.8 %, P = 0.762). However, the former reported a significantly lower proportion of PH assumption assessments compared to the latter (13.6 % versus 42.1 %, P = 0.020). Twenty-five observational studies (25/63) used PSA methods, but two-thirds of them (17/25) showed unclear balance of baseline data after PSA. And the proportion of PH assumption testing after PSA was slightly lower than that before PSA, but the difference was not statistically significant (24.0 % versus 28.0 %, P = 0.317). Comprehensive suggestions were given on confounding control in survival analysis and alternative resolutions for non-compliance with PH assumption. CONCLUSION: This study highlights suboptimal reporting of PH assumption evaluation in observational surgical studies both before and after PSA. Efforts and consensus are needed with respect to the underlying assumptions of statistical methods.

An augmented illness-death model for semi-competing risks with clinically immediate terminal events.

Preeclampsia is a pregnancy-associated condition posing risks of both fetal and maternal mortality and morbidity that can only resolve following delivery and removal of the placenta. Because in its typical form preeclampsia can arise before delivery, but not after, these two events exemplify the time-to-event setting of "semi-competing risks" in which a non-terminal event of interest is subject to the occurrence of a terminal event of interest. The semi-competing risks framework presents a valuable opportunity to simultaneously address two clinically meaningful risk modeling tasks: (i) characterizing risk of developing preeclampsia, and (ii) characterizing time to delivery after onset of preeclampsia. However, some people with preeclampsia deliver immediately upon diagnosis, while others are admitted and monitored for an extended period before giving birth, resulting in two distinct trajectories following the non-terminal event, which we call "clinically immediate" and "non-immediate" terminal events. Though such phenomena arise in many clinical contexts, to-date there have not been methods developed to acknowledge the complex dependencies between such outcomes, nor leverage these phenomena to gain new insight into individualized risk. We address this gap by proposing a novel augmented frailty-based illness-death model with a binary submodel to distinguish risk of immediate terminal event following the non-terminal event. The model admits direct dependence of the terminal event on the non-terminal event through flexible regression specification, as well as indirect dependence via a shared frailty term linking each submodel. We develop an efficient Bayesian sampler for estimation and corresponding model fit metrics, and derive formulae for dynamic risk prediction. In an extended example using pregnancy outcome data from an electronic health record, we demonstrate the proposed model's direct applicability to address a broad range of clinical questions.

Statistical analysis of self-reported health conditions in cohort studies: handling of missing onset age.

OBJECTIVES: This paper discusses methodological challenges in epidemiological association analysis of a time-to-event outcome and hypothesized risk factors, where age/time at the onset of the outcome may be missing in some cases, a condition commonly encountered when the outcome is self-reported. STUDY DESIGN AND SETTING: A cohort study with long-term follow-up for outcome ascertainment such as the Childhood Cancer Survivor Study (CCSS), a large cohort study of 5-year survivors of childhood cancer diagnosed in 1970-1999 in which occurrences and age at onset of various chronic health conditions (CHCs) are self-reported in surveys. Simple methods for handling missing onset age and their potential bias in the exposure-outcome association inference are discussed. The interval-censored method is discussed as a remedy for handling this problem. The finite sample performance of these approaches is compared through Monte Carlo simulations. Examples from the CCSS include four CHCs (diabetes, myocardial infarction, osteoporosis/osteopenia, and growth hormone deficiency). RESULTS: The interval-censored method is useable in practice using the standard statistical software. The simulation study showed that the regression coefficient estimates from the 'Interval censored' method consistently displayed reduced bias and, in most cases, smaller standard deviations, resulting in smaller mean square errors, compared to those from the simple approaches, regardless of the proportion of subjects with an event of interest, the proportion of missing onset age, and the sample size. CONCLUSION: The interval-censored method is a statistically valid and practical approach to the association analysis of self-reported time-to-event data when onset age may be missing. While the simpler approaches that force such data into complete data may enable the standard analytic methods to be applicable, there is considerable loss in both accuracy and precision relative to the interval-censored method.

Longitudinal metabolite and protein trajectories prior to diabetes mellitus diagnosis in Danish blood donors: a nested case-control study.

AIMS/HYPOTHESIS: Metabolic risk factors and plasma biomarkers for diabetes have previously been shown to change prior to a clinical diabetes diagnosis. However, these markers only cover a small subset of molecular biomarkers linked to the disease. In this study, we aimed to profile a more comprehensive set of molecular biomarkers and explore their temporal association with incident diabetes. METHODS: We performed a targeted analysis of 54 proteins and 171 metabolites and lipoprotein particles measured in three sequential samples spanning up to 11 years of follow-up in 324 individuals with incident diabetes and 359 individuals without diabetes in the Danish Blood Donor Study (DBDS) matched for sex and birth year distribution. We used linear mixed-effects models to identify temporal changes before a diabetes diagnosis, either for any incident diabetes diagnosis or for type 1 and type 2 diabetes mellitus diagnoses specifically. We further performed linear and non-linear feature selection, adding 28 polygenic risk scores to the biomarker pool. We tested the time-to-event prediction gain of the biomarkers with the highest variable importance, compared with selected clinical covariates and plasma glucose. RESULTS: We identified two proteins and 16 metabolites and lipoprotein particles whose levels changed temporally before diabetes diagnosis and for which the estimated marginal means were significant after FDR adjustment. Sixteen of these have not previously been described. Additionally, 75 biomarkers were consistently higher or lower in the years before a diabetes diagnosis. We identified a single temporal biomarker for type 1 diabetes, IL-17A/F, a cytokine that is associated with multiple other autoimmune diseases. Inclusion of 12 biomarkers improved the 10-year prediction of a diabetes diagnosis (i.e. the area under the receiver operating curve increased from 0.79 to 0.84), compared with clinical information and plasma glucose alone. CONCLUSIONS/INTERPRETATION: Systemic molecular changes manifest in plasma several years before a diabetes diagnosis. A particular subset of biomarkers shows distinct, time-dependent patterns, offering potential as predictive markers for diabetes onset. Notably, these biomarkers show shared and distinct patterns between type 1 diabetes and type 2 diabetes. After independent replication, our findings may be used to develop new clinical prediction models.

Modeling Recovery Housing Retention and Program Outcomes by Justice Involvement among Residents in Virginia, USA: An Observational Study.

Living in recovery housing can improve addiction recovery and desistance outcomes. This study examined whether retention in recovery housing and types of discharge outcomes (completed, "neutral," and "negative" outcomes) differed for clients with recent criminal legal system (CLS) involvement. Using data from 101 recovery residences certified by the Virginia Association of Recovery Residences based on 1,978 individuals completing the REC-CAP assessment, competing risk analyses (cumulative incidence function, restricted mean survival time, and restricted mean time lost) followed by the marginalization of effects were implemented to examine program outcomes at final discharge. Residents with recent CLS involvement were more likely to be discharged for positive reasons (successful completion of their goals) and premature/negative reasons (e.g., disciplinary releases) than for neutral reasons. Findings indicate that retention for 6-18 months is essential to establish and maintain positive discharge outcomes, and interventions should be developed to enhance retention in recovery residents with recent justice involvement.

Familywise error for multiple time-to-event endpoints in a group sequential design.

We investigate the familywise error rate (FWER) for time-to-event endpoints evaluated using a group sequential design with a hierarchical testing procedure for secondary endpoints. We show that, in this setup, the correlation between the log-rank test statistics at interim and at end of study is not congruent with the canonical correlation derived for normal-distributed endpoints. We show, both theoretically and by simulation, that the correlation also depends on the level of censoring, the hazard rates of the endpoints, and the hazard ratio. To optimize operating characteristics in this complex scenario, we propose a simulation-based method to assess the FWER which, better than the alpha-spending approach, can inform the choice of critical values for testing secondary endpoints.

Bridging the gap between two-stage and joint models: The case of tumor growth inhibition and overall survival models.

Many clinical trials generate both longitudinal biomarker and time-to-event data. We might be interested in their relationship, as in the case of tumor size and overall survival in oncology drug development. Many well-established methods exist for analyzing such data either sequentially (two-stage models) or simultaneously (joint models). Two-stage modeling (2stgM) has been challenged (i) for not acknowledging that biomarkers are endogenous covariable to the survival submodel and (ii) for not propagating the uncertainty of the longitudinal biomarker submodel to the survival submodel. On the other hand, joint modeling (JM), which properly circumvents both problems, has been criticized for being time-consuming, and difficult to use in practice. In this paper, we explore a third approach, referred to as a novel two-stage modeling (N2stgM). This strategy reduces the model complexity without compromising the parameter estimate accuracy. The three approaches (2stgM, JM, and N2stgM) are formulated, and a Bayesian framework is considered for their implementation. Both real and simulated data were used to analyze the performance of such approaches. In all scenarios, our proposal estimated the parameters approximately as JM but without being computationally expensive, while 2stgM produced biased results.

Competing Risks Analysis of Kidney Transplant Waitlist Outcomes: Two Important Statistical Perspectives.

Modern competing risks analysis has 2 primary goals in clinical epidemiology as follows: (i) to maximize the clinician's knowledge of etiologic associations existing between potential predictor variables and various cause-specific outcomes via cause-specific hazard models, and (ii) to maximize the clinician's knowledge of noteworthy differences existing in cause-specific patient risk via cause-specific subdistribution hazard models (cumulative incidence functions [CIFs]). A perfect application exists in analyzing the following 4 distinct outcomes after listing for a deceased donor kidney transplant (DDKT): (i) receiving a DDKT, (ii) receiving a living donor kidney transplant (LDKT), (iii) waitlist removal due to patient mortality or a deteriorating medical condition, and (iv) waitlist removal due to other reasons. It is important to realize that obtaining a complete understanding of subdistribution hazard ratios (HRs) is simply not possible without first having knowledge of the multivariable relationships existing between the potential predictor variables and the cause-specific hazards (perspective #1), because the cause-specific hazards form the "building blocks" of CIFs. In addition, though we believe that a worthy and practical alternative to estimating the median waiting-time-to DDKT is to ask, "what is the conditional probability of the patient receiving a DDKT, given that he or she would not previously experience one of the competing events (known as the cause-specific conditional failure probability)," only an appropriate estimator of this conditional type of cumulative incidence should be used (perspective #2). One suggested estimator, the well-known "one minus Kaplan-Meier" approach (censoring competing events), simply does not represent any probability in the presence of competing risks and will almost always produce biased estimates (thus, it should never be used).

Genetic and early life factors influence on time-to-multiple sclerosis diagnosis: A UK Biobank study.

BACKGROUND: Previous investigations into multiple sclerosis (MS) risk factors predominantly relied on retrospective studies, which do not consider different follow-up times and assume a constant risk effect throughout lifetime. OBJECTIVE: We aimed to evaluate the impact of genetic and early life factors on MS diagnosis by employing a time-to-event analysis in a prospective cohort. METHODS: We used the UK Biobank data, considering the observation period from birth up to 31 December 2022. We considered genetic risk, using a multiple sclerosis polygenic risk score (MS-PRS), and various early life factors. Tobacco smoking and infectious mononucleosis diagnosis were also considered as time-varying variables along the follow-up. Using a Cox proportional hazards model, we examined the associations between these factors and MS diagnosis instantaneous risk. RESULTS: We analyzed 345,027 participants, of which 1669 had an MS diagnosis. Our analysis revealed age-dependent effects for sex (females vs males) and higher MS-PRS, with greater hazard ratios observed in young adults. CONCLUSION: The age-dependent effects suggest that retrospective studies could have underestimated sex and genetic variants' risk roles during younger ages. Therefore, we emphasize the importance of a time-to-event approach using longitudinal data to better characterize age-dependent risk effects.

Predictability of buprenorphine-naloxone treatment retention: A multi-site analysis combining electronic health records and machine learning.

BACKGROUND AND AIMS: Opioid use disorder (OUD) and opioid dependence lead to significant morbidity and mortality, yet treatment retention, crucial for the effectiveness of medications like buprenorphine-naloxone, remains unpredictable. Our objective was to determine the predictability of 6-month retention in buprenorphine-naloxone treatment using electronic health record (EHR) data from diverse clinical settings and to identify key predictors. DESIGN: This retrospective observational study developed and validated machine learning-based clinical risk prediction models using EHR data. SETTING AND CASES: Data were sourced from Stanford University's healthcare system and Holmusk's NeuroBlu database, reflecting a wide range of healthcare settings. The study analyzed 1800 Stanford and 7957 NeuroBlu treatment encounters from 2008 to 2023 and from 2003 to 2023, respectively. MEASUREMENTS: Predict continuous prescription of buprenorphine-naloxone for at least 6 months, without a gap of more than 30 days. The performance of machine learning prediction models was assessed by area under receiver operating characteristic (ROC-AUC) analysis as well as precision, recall and calibration. To further validate our approach's clinical applicability, we conducted two secondary analyses: a time-to-event analysis on a single site to estimate the duration of buprenorphine-naloxone treatment continuity evaluated by the C-index and a comparative evaluation against predictions made by three human clinical experts. FINDINGS: Attrition rates at 6 months were 58% (NeuroBlu) and 61% (Stanford). Prediction models trained and internally validated on NeuroBlu data achieved ROC-AUCs up to 75.8 (95% confidence interval [CI] = 73.6-78.0). Addiction medicine specialists' predictions show a ROC-AUC of 67.8 (95% CI = 50.4-85.2). Time-to-event analysis on Stanford data indicated a median treatment retention time of 65 days, with random survival forest model achieving an average C-index of 65.9. The top predictor of treatment retention identified included the diagnosis of opioid dependence. CONCLUSIONS: US patients with opioid use disorder or opioid dependence treated with buprenorphine-naloxone prescriptions appear to have a high (∼60%) treatment attrition by 6 months. Machine learning models trained on diverse electronic health record datasets appear to be able to predict treatment continuity with accuracy comparable to that of clinical experts.

Lost in the plot: missing visual elements in Kaplan-Meier plots of phase III oncology trials.

Missing visual elements (MVE) in Kaplan-Meier (KM) curves can misrepresent data, preclude curve reconstruction, and hamper transparency. This study evaluated KM plots of phase III oncology trials. MVE were defined as an incomplete y-axis range or missing number at risk table in a KM curve. Surrogate endpoint KM curves were additionally evaluated for complete interpretability, defined by (1) reporting the number of censored patients and (2) correspondence of the disease assessment interval with the number at risk interval. Among 641 trials enrolling 518 235 patients, 116 trials (18%) had MVE in KM curves. Industry sponsorship, larger trials, and more recently published trials were correlated with lower odds of MVE. Only 3% of trials (15 of 574) published surrogate endpoint KM plots with complete interpretability. Improvements in the quality of KM curves of phase III oncology trials, particularly for surrogate endpoints, are needed for greater interpretability, reproducibility, and transparency in oncology research.