Recurrent Event Models Research Articles

Reduction in shock burden with catheter ablation versus escalated antiarrhythmic drug therapy: Insights from the VANISH trial

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): 1. Canadian Institutes of Health Research (CIHR) 2. Additional financial support from St. Jude Medical and Biosense Webster Background Recurrent shocks for ventricular tachycardia (VT) are associated with an increased risk of heart failure and mortality and have a negative influence on quality of life. Catheter ablation has been shown to improve VT event-free survival in patients with antiarrhythmic drug (AAD)-refractory VT and prior myocardial infarction (MI); however, the effects of ablation on shock burden has yet to be investigated. Purpose Our primary objectives were to compare the shock-treated VT event burden and appropriate shock burden following randomization to treatment with either catheter ablation or escalated AAD therapy among VT patients with prior MI in the Ventricular tachycardia AblatioN versus escalated antiarrhythmic drug therapy in Ischemic Heart disease (VANISH) randomized trial. Methods Recurrent event analyses were performed using the intention-to-treat population of the VANISH trial. Shock-treated VT event burden was defined as the total number of VT events treated with ≥1 appropriate internal or external shocks. Appropriate shock burden was defined as the total number of appropriate internal and external shocks delivered, regardless of the number of VT events. All VT events and implantable cardioverter defibrillator (ICD) therapies were adjudicated by reviewers blinded to the treatment allocation. Three recurrent event models were used to compare the shock burden between treatment arms (Anderson-Gill (AG), Frailty, and Prentice, Williams, and Peterson Total Time (PWP-TT). Each model clustered by patient and accounted for competing risk of death with the Fine and Gray sub-distributions hazards model. Results Of the 259 patients enrolled in the VANISH trial [median age 69.8 (IQR 63.0-74.2) years, 7.0% women], 132 patients were randomized to ablation and 129 patients to escalated AAD therapy. Over a median follow-up of 23.4 (IQR 14.7-40.4) months, there were 138 shock-treated VT events [39.07 (95% CI 33.14-46.07) shock-treated VT events per 100 person-years] in the ablation arm and 218 shock-treated VT events [64.60 (95% CI 56.49–73.84) shock-treated VT events per 100 person-years] in the escalated AAD therapy arm (Figure 1). Ablation patients had a 40% lower shock-treated VT event burden (ie. number of shock-treated VT events) compared to patients randomized to escalated AAD therapy [Figure 1; AG HR 0.60 (95% 0.38-0.95)]. Further, there was also a statistically significant reduction in the appropriate shock burden (i.e. number of appropriate shocks) among ablation patients (169 appropriate shocks) compared to escalated AAD therapy patients (266 appropriate shocks) [Figure 1; AG HR 0.61 (95% CI 0.37-0.96)]. All results were consistent between the 3 recurrent event models. Conclusion Among patients with AAD-refractory VT and a prior MI, catheter ablation reduced shock-treated VT event burden and appropriate shock burden compared to escalated AAD therapy.

C-index regression for recurrent event data

Recurrent event data analysis plays an important role in many fields, e.g., medicine, social science, and economics. While the existing approaches under the proportional rates or mean model yield poor performance when the underlying model is misspecified, we propose a novel model-free approach by introducing a lower bound on the concordance index (C-Index). We develop an estimation method through deriving a continuous lower bound on the C-Index based on the log-sigmoid function and also provide a variable selection procedure in high dimensional settings. Under both low and high dimensional settings, simulation results show that the proposed methods outperform the gamma frailty recurrent event model when the proportional mean assumption is violated. Moreover, an application to the hospital readmission dataset shows results in line with previous studies and a higher C-Index value further assures model decency.

The bleeding phenotype in people with nonsevere hemophilia

AbstractDetailed information on the onset, frequency, and severity of bleeding in nonsevere hemophilia is limited. We aimed to assess the bleeding phenotype of persons with nonsevere hemophilia and to analyze the association between baseline factor VIII/IX (FVIII/IX) levels and the joint bleeding rate. In the DYNAMO (Dynamic Interplay Between Bleeding Phenotype and Baseline Factor Level in Moderate and Mild Hemophilia A and B) study, an international multicenter cohort, we included males with nonsevere hemophilia (FVIII/IX, 0.02-0.35 IU/mL) aged 12 to 55 years. Information on age at first treated (joint) bleed, annual bleeding rates (ABRs), and annual joint bleeding rates (AJBRs) was collected from the medical files. The association between baseline FVIII/IX levels and the joint bleeding rate was assessed by using a frailty model for recurrent events. In total, 304 persons (70 with moderate hemophilia and 234 with mild hemophilia) were included. The median age was 38 years (interquartile range [IQR], 25-49 years), and the median baseline FVIII/IX level was 0.12 IU/mL (IQR, 0.05-0.21 IU/mL). In total, 245 (81%) persons had experienced at least 1 bleed, and 156 (51%) had experienced at least 1 joint bleed. The median age at first bleed and first joint bleed was 8 and 10 years, respectively. The median ABR and AJBR was 0.2 (IQR, 0.1-0.5) and 0.0 (IQR, 0.0-0.2). From baseline FVIII/IX levels 0.02 to 0.05 IU/mL to >0.25 IU/mL, the median ABR decreased from 0.6 (IQR, 0.2-1.4) to 0.1 (IQR, 0.0-0.2) and the AJBR from 0.2 (IQR, 0.0-0.4) to 0.0 (IQR, 0.0-0.0). Baseline FVIII/IX was inversely associated with the joint bleeding rate (P < .001). Low bleeding rates were observed in persons with nonsevere hemophilia. However, one-half of all adolescents and adults had experienced a joint bleed.

Estimation of recurrent atherosclerotic cardiovascular event risk in patients with established cardiovascular disease: the updated SMART2 algorithm.

AimsThe 10-year risk of recurrent atherosclerotic cardiovascular disease (ASCVD) events in patients with established ASCVD can be estimated with the Secondary Manifestations of ARTerial disease (SMART) risk score, and may help refine clinical management. To broaden generalizability across regions, we updated the existing tool (SMART2 risk score) and recalibrated it with regional incidence rates and assessed its performance in external populations.Methods and resultsIndividuals with coronary artery disease, cerebrovascular disease, peripheral artery disease, or abdominal aortic aneurysms were included from the Utrecht Cardiovascular Cohort-SMART cohort [n = 8355; 1706 ASCVD events during a median follow-up of 8.2 years (interquartile range 4.2–12.5)] to derive a 10-year risk prediction model for recurrent ASCVD events (non-fatal myocardial infarction, non-fatal stroke, or cardiovascular mortality) using a Fine and Gray competing risk-adjusted model. The model was recalibrated to four regions across Europe, and to Asia (excluding Japan), Japan, Australia, North America, and Latin America using contemporary cohort data from each target region. External validation used data from seven cohorts [Clinical Practice Research Datalink, SWEDEHEART, the international REduction of Atherothrombosis for Continued Health (REACH) Registry, Estonian Biobank, Spanish Biomarkers in Acute Coronary Syndrome and Biomarkers in Acute Myocardial Infarction (BACS/BAMI), the Norwegian COgnitive Impairment After STroke, and Bialystok PLUS/Polaspire] and included 369 044 individuals with established ASCVD of whom 62 807 experienced an ASCVD event. C-statistics ranged from 0.605 [95% confidence interval (CI) 0.547–0.664] in BACS/BAMI to 0.772 (95% CI 0.659–0.886) in REACH Europe high-risk region. The clinical utility of the model was demonstrated across a range of clinically relevant treatment thresholds for intensified treatment options.ConclusionThe SMART2 risk score provides an updated, validated tool for the prediction of recurrent ASCVD events in patients with established ASCVD across European and non-European populations. The use of this tool could allow for a more personalized approach to secondary prevention based upon quantitative rather than qualitative estimates of residual risk.

Cluster analysis of patients with granulomatosis with polyangiitis (GPA) based on clinical presentation symptoms: a UK population-based cohort study

BackgroundGranulomatosis with polyangiitis (GPA) is small vessel vasculitis with heterogeneous clinical presentation. In the present population-based cohort study, we classified patients with GPA based on clinical features at presentation using an unsupervised clustering approach and compared their mortality, infections and frequency of comorbidities.MethodsIn this open cohort study, de-identified primary care data of patients with GPA included in the IQVIA Medical Research Data database between 1 January 1995 and 25 September 2019 was analysed retrospectively. Latent class analysis was performed to create symptom clusters of patients based on 16 categories of symptoms representing various organ involvement. All-cause mortality of resultant clusters was compared after adjusting for age, sex, Townsend deprivation quintile and smoking status at index date using extended Cox proportional hazards models. Prescription of antibiotics, considered as an indirect indicator of recurrent bacterial infection, was compared using a recurrent event model, after adjusting for quarterly use of steroid as a time-dependent covariate. Cumulative frequencies of common comorbidities were compared among the clusters at index visit, 1-year and 3-year follow-up.ResultsAltogether, 649 patients with GPA [median age 60.0 (IQR: 49.6–70.1)] were included. Three clusters were identified: patients with limited disease mainly with involvement of ENT and cough were classified into cluster 1 (n = 426); cluster 2 had generalised non-renal disease (n = 176); while patients in cluster 3 had renal-predominant disease (n = 47). Many patients in cluster 1 developed generalised disease at the end of 1 year. Mortality in clusters 2 and 3 was higher compared with cluster 1. Mortality in cluster 1 itself was 68% higher than the general population without GPA. The duration of antibiotics prescription and frequency of coexisting medical illnesses was also higher in clusters 2 and 3.ConclusionsIn a primary care setting, patients with GPA can be classified into three distinct clusters with different prognosis, susceptibility to recurrent infections and presence of comorbidities. The tendency of cluster 1 to evolve into a more generalised disease raises questions about current immunosuppressive treatment approaches in these patients.

A review of effective age models and associated non- and semiparametric methods

First an overview of a class of models for recurrent events is given. The class of models considered is known as virtual or effective age models. One of the strengths of this class of models is their ability to account for intervention effects after an event occurrence. Some of the models within this class allow to account for the effects of covariates and the impact of the number of already observed events. After having provided an overview of this class of models, non- and semiparametric inference methods for these models are reviewed. Several open problems in non- and semiparametric inference methods for these models are also described.

An Introspective Overview of the Dynamics of Recurrent Events Data Analysis

A recurrent event remains the outcome variable of interest in many biometric studies. Recurrent events can be explained as events of defined interest that can occur to same person more than once during the study period. This study presents an overview of different pertinent recurrent models for analyzing recurrent events.&#x0D; Aims: To introduce, compare, evaluate and discuss pros and cons of four models in analyzing recurrent events, so as to validate previous findings in respect of the superiority or appropriateness of these models.&#x0D; Study Design: A comparative studies based on simulation of recurrent event models applied to a tertiary data on cancer studies. &#x0D; Methodology: Codes in R were implemented for simulating four recurrent event models, namely; The Andersen and Gill model; Prentice, Williams and Peterson models; Wei, Lin and Weissferd; and Cox frailty model. Finally, these models were applied to analyze the first forty subjects from a study of Bladder Cancer Tumors. The data set contained the first four repetitions of the tumor for each patient, and each recurrence time was recorded from the entry time of the patient into the study. An isolated risk interval is defined by each time to an event or censoring.&#x0D; Results: The choice and usage of any of the models lead to different conclusions, but the choice depends on: risk intervals; baseline hazard; risk set; and correlation adjustment or simplistically, type of data and research question. The PWP-GT model could be used if the research question is focused on whether treatment was effective for the event since the previous event happened. However, if the research question is designed to find out whether treatment was effective for the event since the start of treatment, then we could use the PWP- TT. The AG model will be adequate if a common baseline hazard could be assumed, but the model lacks the details and versatility of the event-specific models. The WLW model is very suitable for data with diverse events for the same person, which underscores a potentially different baseline hazard for each type.&#x0D; Conclusion: PWP-GT has proven to be the most useful model for analyzing recurrent event data.

Factors involved in time reduction between seizure relapses in patients with epilepsy attending emergency rooms in Medellín, Colombia

BackgroundSeizure relapses are the leading cause of admission to emergency rooms (ER) in people with epilepsy. ObjectiveTo analyze administrative and clinical factors associated with the duration between seizure relapses in people with epilepsy admitted to the Neurological Institute of Colombia (Medellin) between July 2018 and July 2019. Materials and methodsA retrospective follow-up study of 156 patients over 18 years old, diagnosed with epilepsy, and treated for over a year. The outcome variable was the time between seizure relapses, identified through the record of ER attendances. In addition, difficulties in the prescription filling process (delay, omission, or brand change) and clinical characteristics were analyzed as potential associated influence factors. The statistical analysis was performed using the Prentice, Williams & Peterson-Gap Time survival model for recurrent events. Finally, Adjusted Hazard Ratios (aHR) with 95% confidence intervals (95%CI) are also presented. ResultsOne hundred fifty-six patients were analyzed. Their average age of diagnosis was 15.5 years (SD = 22.5), the median number of monthly seizures was 3 (SD = 9.3), and 50.6% were women. Moreover, difficulties in the prescription filling process were associated with a time reduction between seizure relapses (aHR = 2.61; 95%CI 1.49–4.57), showing a similar impact as having a history of three or four types of events (aHR = 2.96; 95%CI 1.23–7.12) and neuropsychiatric comorbidity (aHR = 1.89; 95%CI 1.04–3.54). ConclusionNeuropsychiatric comorbidity, history of several types of events, and experiencing difficulties with prescription filling are associated with lower benefit from treatment to control seizure relapses.

Attainment of low disease activity and remission targets reduces the risk of severe flare and new damage in childhood lupus.

ObjectivesTo assess the achievability and effect of attaining low disease activity (LDA) or remission in childhood-onset SLE (cSLE).MethodsAttainment of three adult-SLE derived definitions of LDA (LLDAS, LA, Toronto-LDA), and four definitions of remission (clinical-SLEDAI-defined remission on/off treatment, pBILAG-defined remission on/off treatment) was assessed in UK JSLE Cohort Study patients longitudinally. Prentice–Williams–Petersen gap recurrent event models assessed the impact of LDA/remission attainment on severe flare/new damage.ResultsLLDAS, LA and Toronto-LDA targets were reached in 67%, 73% and 32% of patients, after a median of 18, 15 or 17 months, respectively. Cumulatively, LLDAS, LA and Toronto-LDA was attained for a median of 23%, 31% and 19% of total follow-up-time, respectively. Remission on-treatment was more common (61% cSLEDAI-defined, 42% pBILAG-defined) than remission off-treatment (31% cSLEDAI-defined, 21% pBILAG-defined). Attainment of all target states, and disease duration (>1 year), significantly reduced the hazard of severe flare (P < 0.001). As cumulative time in each target increased, hazard of severe flare progressively reduced. LLDAS attainment reduced the hazard of severe flare more than LA or Toronto-LDA (P < 0.001). Attainment of LLDAS and all remission definitions led to a statistically comparable reduction in the hazards of severe flare (P > 0.05). Attainment of all targets reduced the hazards of new damage (P < 0.05).ConclusionsThis is the first study demonstrating that adult-SLE-derived definitions of LDA/remission are achievable in cSLE, significantly reducing risk of severe flare/new damage. Of the LDA definitions, LLDAS performed best, leading to a statistically comparable reduction in the hazards of severe flare to attainment of clinical remission.

Periodic screening for breast and cervical cancer in women with diabetes: a population-based cohort study.

Diabetes is associated with poorer cancer outcomes. Screening for breast and cervical cancer is recommended by clinical guidelines; however, utilization of these tests in people with diabetes has been unclear due to methodological limitations in the evidence base. We used administrative data to determine the association between diabetes and the rates of becoming up-to-date with periodic breast and cervical cancer screening over a 20-year period. Healthcare databases from Ontario, Canada, were linked to assemble two population-based cohorts of 50-70 and 21-70year-olds between 1994 and 2011, eligible for breast and cervical cancer screening, respectively. Using age as the time scale, multivariable recurrent events models were implemented to examine the association between the presence of diabetes and the rates of becoming up-to-date with the recommended cancer screenings. In each of the breast and cervical cancer screening cohorts, there were, respectively, 1,516,302 (16% had diabetes at baseline) and 4,751,220 (9.5% had diabetes at baseline) screen-eligible women. In multivariable models, prevalent diabetes (duration ≥ 2years) was associated with lower rates of becoming up-to-date with cervical (hazard ratio, HR 0.85, 95% confidence interval, CI 0.84-0.85) and breast (HR 0.94, CI 0.93-0.94) cancer screening, compared to no diabetes. Having diabetes is associated with decreased rates of becoming up-to-date with two recommended periodic cancer screenings, with a bigger reduction in the rates of becoming up-to-date with cervical cancer screening. Greater attention to cervical cancer preventive services is needed in women with diabetes.

Morbidity and health care use among siblings of children with cancer: A population-based study.

Childhood cancer impacts the entire family unit. We sought to investigate its impact on the long-term physical health outcomes of siblings of children with cancer. Pediatric cancer patients diagnosed in Ontario, Canada between 1988 and 2016 were linked to biological siblings. Sibling cases were matched to population controls based on sex, age, geographic location, and number of other children in the family. After individual linkage to health services data, we compared several outcomes between sibling cases and controls: (a) physical health conditions (such as diabetes, hypertension, and death); (b) acute health care use (hospitalization, low- and high-acuity emergency department [ED] visits); and (c) preventive health care use (periodic health checkups, influenza vaccinations). Cox proportional hazards, recurrent event, or logistic regression models were used as appropriate. We identified 8529 sibling cases and 30,364 matched controls (median age at index: 6years, median age at last follow-up 17years). Compared to controls, siblings were at increased risk of hypertension (hazard ratio [HR] 1.8; 95% confidence interval [CI] 1.1-2.9; p=.01), had higher rates of low- and high-acuity ED visits (rate ratio 1.1; 95% CI 1.1-1.2; p<.001), and increased risk of hospitalization (HR 1.1; 95% CI 1.1-1.2; p<.001). Sibling cases were also more likely to receive preventive health care (p<.05). Increased risk of hypertension, high-acuity ED visits, and hospitalizations suggest that siblings may experience poorer health compared to controls. Counseling families about this potential increased risk and long-term follow-up of siblings to monitor their physical health may be justified.

Bivariate pseudo-observations for recurrent event analysis with terminal events.

The analysis of recurrent events in the presence of terminal events requires special attention. Several approaches have been suggested for such analyses either using intensity models or marginal models. When analysing treatment effects on recurrent events in controlled trials, special attention should be paid to competing deaths and their impact on interpretation. This paper proposes a method that formulates a marginal model for recurrent events and terminal events simultaneously. Estimation is based on pseudo-observations for both the expected number of events and survival probabilities. Various relevant hypothesis tests in the framework are explored. Theoretical derivations and simulation studies are conducted to investigate the behaviour of the method. The method is applied to two real data examples. The bivariate marginal pseudo-observation model carries the strength of a two-dimensional modelling procedure and performs well in comparison with available models. Finally, an extension to a three-dimensional model, which decomposes the terminal event per death cause, is proposed and exemplified.

Hearing Loss After Radiation and Chemotherapy for Central Nervous System and Head and Neck Tumors in Children

Hearing loss (HL) is a serious secondary effect of treatment for head and neck (H&N) and central nervous system (CNS) tumors in children. Radiation and platinum chemotherapy independently increase the risk of HL; however, combined modality treatment is routinely used and the effect of chemoradiation on HL risk is not well-studied. Using chemotherapy and cochlear radiation (RT) doses, we created a model that predicts for HL, including a nomogram that can be used in the clinical setting to calculate the risk of clinically-significant HL.In this single institution retrospective study, 171 patients with H&N or CNS tumors were treated with radiation, with or without chemotherapy and had longitudinal (≥2) audiological evaluation. SIOP-Boston (SIOP) and Chang grades were assigned to 2,420 hearing assessments of 342 ears; analyses using SIOP grades are presented here. For the nomogram, SIOP grade ≥3 was considered clinically-significant HL (requiring hearing aids). An Andersen-Gill model for recurrent events was used to obtain inverse intensity weights to account for factors explaining the number of assessments. Multivariable weighted ordinal logistic regression was fitted to evaluate the effect of clinicopathologic features on HL. Clustered robust standard errors were calculated to account for intra-patient correlation.Patients underwent a median of 6 (range 2-23) assessments over a median of 3.1 years from diagnosis to last audiogram (range 0.1-15.2 years). Cisplatin was given to 63% of patients and 34% received carboplatin. The mean cochlear doses on the right and left were 36.8 Gy (standard deviation [SD] 16.5) and 37.0 Gy (SD 16.2), respectively. Multivariable regression revealed that mean cochlear dose (odds ratio [OR] 1.04 per Gy, 95% confidence interval [CI] 1.02-1.05, P < 0.001), time since RT (OR 1.2 per year, 95% CI 1.2-1.3, P < 0.001), cisplatin use (OR 5.33, 95% CI 2.9-9.9, P < 0.001), and carboplatin use (OR 2.3, 95% CI 1.27-4.17, P = 0.006) were associated with increasing SIOP grade of HL; age at RT, hydrocephalus, surgery, amifostine, and laterality were not correlated with HL. There was no synergistic effect of RT and cisplatin (interaction term, P = 0.4) or RT and carboplatin (interaction term, P = 0.9). Cumulative incidence of high-frequency HL (> 4 kHz) was > 50% at 5 years post-RT in those who received mean dose > 20 Gy to the cochlea, while incidence of HL across all frequencies continued to increase beyond 5 years post-RT.RT and chemotherapy have an additive, not synergistic, effect on HL risk. Mean cochlear radiation dose, time since radiation, and platinum chemotherapy use were associated with HL. This modelling can guide survivorship care by multidisciplinary pediatric oncology teams with respect to audiology follow-up in an effort to that ensure suitable educational accommodations and assistive devices are in place.

Prediction of recurrent stroke among ischemic stroke patients with atrial fibrillation: Development and validation of a risk score model.

There is currently no validated risk prediction model for recurrent events among patients with acute ischemic stroke (AIS) and atrial fibrillation (AF). Considering that the application of conventional risk scores has contextual limitations, new strategies are needed to develop such a model. Here, we set out to develop and validate a comprehensive risk prediction model for stroke recurrence in AIS patients with AF. AIS patients with AF were collected from multicenter registries in South Korea and Japan. A developmental dataset was constructed with 5648 registered cases from both countries for the period 2011‒2014. An external validation dataset was also created, consisting of Korean AIS subjects with AF registered between 2015 and 2018. Event outcomes were collected during 1 year after the index stroke. A multivariable prediction model was developed using the Fine-Gray subdistribution hazard model with non-stroke mortality as a competing risk. The model incorporated 21 clinical variables and was further validated, calibrated, and revised using the external validation dataset. The developmental dataset consisted of 4483 Korean and 1165 Japanese patients (mean age, 74.3 ± 10.2 years; male 53%); 338 patients (6%) had recurrent stroke and 903 (16%) died. The clinical profiles of the external validation set (n = 3668) were comparable to those of the developmental dataset. The c-statistics of the final model was 0.68 (95% confidence interval, 0.66 ‒0.71). The developed prediction model did not show better discriminative ability for predicting stroke recurrence than the conventional risk prediction tools (CHADS2, CHA2DS2-VASc, and ATRIA). Neither conventional risk stratification tools nor our newly developed comprehensive prediction model using available clinical factors seemed to be suitable for identifying patients at high risk of recurrent ischemic stroke among AIS patients with AF in this modern direct oral anticoagulant era. Detailed individual information, including imaging, may be warranted to build a more robust and precise risk prediction model for stroke survivors with AF.

An application of Markov chain modeling and semi-parametric regression for recurrent events in health data

Longitudinal studies are best suited to describe the evolution of particular health conditions over time. In this study, data on the occurrence and transition of cough among post-operative cardiac surgery patients was analyzed using semi-parametric regression models for recurrent events. Cough severity was recorded as no cough, mild cough and severe cough. Transition probability matrix was calculated for the various transitions and across different covariate categories. Also, mean first passage time (MFPT) was calculated using Markov principles and Monte-Carlo simulation. The Andersen-Gill (AG) and Prentice, Williams and Peterson (PWP) semi-parametric regression models were used to test the effect of covariates on the cough transition. Ninety percent of the patients developed cough on the first post-operative day. The probability of transitioning from no cough to severe cough was 8% but the probability of resolution was just 3%. The mean first passage time from no cough to severe cough was about 7.2 (95% CI 6.8–7.5) days and the resolution time was 13.7 (13.0–14.5) days. The MFPT varied widely across the covariate categories. The regression models did not reveal any major significant influences by the measured covariates and the models without covariates were not significantly different from the covariate models. Applying these statistical techniques can serve as effective tools to help medical decision makers to provide better, consistent, efficient and evidence-based healthcare services.

A comparison of semiparametric approaches to evaluate composite endpoints in heart failure trials

In heart failure (HF) trials efficacy is usually assessed by a composite endpoint including cardiovascular death (CVD) and heart failure hospitalizations (HFHs), which has traditionally been evaluated with a time-to-first-event analysis based on a Cox model. As a considerable fraction of events is ignored that way, methods for recurrent events were suggested, among others the semiparametric proportional rates models by Lin, Wei, Yang, and Ying (LWYY model) and Mao and Lin (Mao-Lin model). In our work we apply least false parameter theory to explain the behavior of the composite treatment effect estimates resulting from the Cox model, the LWYY model, and the Mao-Lin model in clinically relevant scenarios parameterized through joint frailty models. These account for both different treatment effects on the two outcomes (CVD, HFHs) and the positive correlation between their risk rates. For the important setting of beneficial outcome-specific treatment effects we show that the correlation results in composite treatment effect estimates, which are decreasing with trial duration. The estimate from the Cox model is affected more by the attenuation than the estimates from the recurrent event models, which both demonstrate very similar behavior. Since the Mao-Lin model turns out to be less sensitive to harmful effects on mortality, we conclude that, among the three investigated approaches, the LWYY model is the most appropriate one for the composite endpoint in HF trials. Our investigations are motivated and compared with empirical results from the PARADIGM-HF trial (ClinicalTrials.gov identifier: NCT01035255), a large multicenter trial including 8399 chronic HF patients.

A/B-Test of Retention and Monetization Using the Cox Model

A/B testing is a popular tool for guiding mobile game development. The developer releases different versions of a game to different test cohorts, and observes which version has the best player retention or monetization. Correctly determining whether the differences are statistically significant is however challenging. Typically the analysis needs to be done on small and heterogeneous player cohorts, with differing follow-up times and unknown player churn. In this paper, we show for the first time how these issues can be properly addressed using the Cox model for recurrent events. The method enables a multivariate A/B-test, that allows determining which game version has the highest player retention or purchase rate, with confidence intervals provided. We demonstrate the benefits of the approach in multiple game development problems, on real-world free-to-play mobile game data.

A varying-coefficient model for gap times between recurrent events.

Recurrent events often arise in follow-up studies where a subject may experience multiple occurrences of the same type of event. Most regression models for recurrent events consider the time scale measured from the study origin and assume constant effects of covariates. In many applications, however, gap times between recurrent events are of natural interest and moreover the effects may actually vary over time. In this article, we propose a marginal varying-coefficient model for gap times between recurrent events that allows for the intra-individual correlation between events. Estimation and inference procedures are developed for the varying coefficients. Consistency and weak convergence of the proposed estimator are established. Monte Carlo simulation studies demonstrate that the proposed method works well with practical sample sizes. The proposed method is illustrated with an analysis of bladder tumor clinical data.

Reconstructing spatial information diffusion networks with heterogeneous agents and text contents

AbstractIt is important to reconstruct the hidden network structure from the infection status change of an information propagation process for evidence‐based spatial decision‐making. Unlike previous work, we not only consider the heterogeneity of the propagation agents, but also incorporate the heterogeneity of the text contents of information within the propagation process. In addition, the infection status is no longer restricted to the binary type (infected or not), and we allow the number of pieces of information texts to be counted which represents the degree of infection. The resulting model is a network‐based multivariate recurrent event model, in which the interactions between different types of text, between different agents, between agents and text types, and their mutual impacts on the whole propagation process can be comprehensively investigated. On that basis, a nonparametric mean‐field equation is derived to govern the propagation process, and a compressive sensing algorithm is provided to infer the hidden spatial propagation network from the infection status data. Finally, the proposed methodology is tested through synthetic data and a real data set of information diffusion on Twitter.

Proton pump inhibitor use is associated with increased rates of post-TIPS hepatic encephalopathy: Replication in an independent patient cohort

PurposeProton pump inhibitor (PPI) use is a potential risk factor for hepatic encephalopathy (HE), but few studies have examined the effect on post-TIPS HE. The purpose of this study was to determine whether PPIs are associated with increased rates of post-TIPS HE in an independent patient cohort. Materials and methodsThis single-institution retrospective study analyzed 86 patients (54 male, mean age 58.2) following TIPS from 1/1/2017 to 12/31/2019. Dates of PPI usage and episodes of new or worsening HE were recorded. Poisson regression with generalized estimating equations was used to test for association between PPI use and post-TIPS HE and to test for dose dependence. Post-TIPS HE was also analyzed using the Andersen-Gill survival model for recurrent events. ResultsThere were 1.88 episodes of new or worsening post-TIPS HE per person-year among 35 patients on uninterrupted PPIs therapy, 1.95 on PPIs and 0.94 off PPIs among 35 patients on intermittent therapy, and 0.47 among 16 patients never on PPIs. PPI use was significantly associated with post-TIPS HE in both univariable (incidence rate ratio (IRR) = 2.62; CI = 1.41–4.84; p = 0.002) and multivariable (IRR = 2.31; CI = 1.37–3.89; p = 0.002) regression. Analysis of only those patients on PPIs showed increased rates of HE with higher doses (IRR = 1.17 per 10 mg omeprazole equivalent; CI = 1.04–1.33; p = 0.011). Recurrent events survival analysis supported the association between PPI use and HE in univariable (hazard ratio (HR) = 2.17; CI = 1.19–3.95; p = 0.011) and multivariable (HR = 1.87; CI = 1.12–3.13; p = 0.017) analysis. ConclusionIn an independent patient cohort PPI use was associated with increased rates of new or worsening post-TIPS HE.