Analysis Of Recurrent Events Research Articles

Bayesian Semiparametric Joint Regression Analysis of Recurrent Adverse Events and Survival in Esophageal Cancer Patients.

We propose a Bayesian semiparametric joint regression model for a recurrent event process and survival time. Assuming independent latent subject frailties, we define marginal models for the recurrent event process intensity and survival distribution as functions of the subject's frailty and baseline covariates. A robust Bayesian model, called Joint-DP, is obtained by assuming a Dirichlet process for the frailty distribution. We present a simulation study that compares posterior estimates under the Joint-DP model to a Bayesian joint model with lognormal frailties, a frequentist joint model, and marginal models for either the recurrent event process or survival time. The simulations show that the Joint-DP model does a good job of correcting for treatment assignment bias, and has favorable estimation reliability and accuracy compared with the alternative models. The Joint-DP model is applied to analyze an observational dataset from esophageal cancer patients treated with chemo-radiation, including the times of recurrent effusions of fluid to the heart or lungs, survival time, prognostic covariates, and radiation therapy modality.

Generalized frailty models for analysis of recurrent events

In this paper, we propose an extension of the frailty modeling to analyze the recurrent events data. We proposed a class of models, based on nonlinear mixed effects modeling, that takes into consideration of the between subject heterogeneity in the model. We propose an estimation method for estimating parameters of the proposed model and their standard deviations. The estimating equations are shown to give consistent estimates under commonly satisfied regularity conditions. A method for consistently estimating the covariance matrix of between subject random effects is also given. We prove the consistency and asymptotical normality of the estimators. We apply the proposed generalized frailty model to the Mammary Tumor data of 48 rats that was previously analyzed by Gail et al. (1980) and Cook and Lawless (2007). We compare the result of the our proposed model with results from the commonly used fixed effect model and frailty model. Both fixed effects model and frailty models result in biased estimates of the variance function for cumulative number of tumors. Simulation results suggest that the proposed generalized frailty model and estimation method is computationally feasible and provides unbiased estimates of the parameters and standard errors of their estimators.

Momentum or market? Determinants of large stock price changes in an emerging market

ABSTRACTThe aim of this study is to investigate the determinants of large price changes in Turkey. We also provide additional evidence on determinants of large price changes in different macroeconomic environments, specifically on the pre-crisis and post-crisis periods. Using recurrent event analysis with stratified observations and frailty effects, our findings suggest that momentum has a significant impact on large price changes during both pre-crisis and post-crisis periods. However, the impact of market is more significant on the estimation of large price declines in the pre-crisis period and of large price increases in the post-crisis period. Additional findings suggest that liquidity and market-to-book ratio have positive, firm size has a negative impact on likelihood of large price changes regardless of the direction of the stock price change and macroeconomic environment. Findings of this study provide new insights into the understanding of large price changes in an emerging market.

The Econometric Analysis of Recurrent Events in Macroeconomics and Finance, by DonHarding and AdrianPagan (Princeton University Press, Princeton, NJ, 2016).

The Econometric Analysis of Recurrent Events in Macroeconomics and Finance, by DonHarding and AdrianPagan (Princeton University Press, Princeton, NJ, 2016).

Non-parametric recurrent events analysis with BART and an application to the hospital admissions of patients with diabetes

Much of survival analysis is concerned with absorbing events, i.e., subjects can only experience a single event such as mortality. This article is focused on non-absorbing or recurrent events, i.e., subjects are capable of experiencing multiple events. Recurrent events have been studied by many; however, most rely on the restrictive assumptions of linearity and proportionality. We propose a new method for analyzing recurrent events with Bayesian Additive Regression Trees (BART) avoiding such restrictive assumptions. We explore this new method via a motivating example of hospital admissions for diabetes patients and simulated data sets.

Semiparametric varying-coefficient regression analysis of recurrent events with applications to treatment switching.

This paper investigates the semiparametric statistical methods for recurrent events. The mean number of the recurrent events are modeled with the generalized semiparametric varying-coefficient model that can flexibly model three types of covariate effects: time-constant effects, time-varying effects, and covariate-varying effects. We assume that the time-varying effects are unspecified functions of time and the covariate-varying effects are parametric functions of an exposure variable specified up to a finite number of unknown parameters. Different link functions can be selected to provide a rich family of models for recurrent events data. The profile estimation methods are developed for the parametric and nonparametric components. The asymptotic properties are established. We also develop some hypothesis testing procedures to test validity of the parametric forms of covariate-varying effects. The simulation study shows that both estimation and hypothesis testing procedures perform well. The proposed method is applied to analyze a data set from an acyclovir study and investigate whether acyclovir treatment reduces the mean relapse recurrences.

Shortening the washout to 4 weeks when switching from natalizumab to fingolimod and risk of disease reactivation in multiple sclerosis

BACKGROUNDThere is limited evidence about the optimal length of washout when switching from natalizumab to fingolimod. OBJECTIVETo study if a washout period of 4 weeks is associated with less disease activity compared to 8 weeks. METHODS25 patients with Relapsing Remitting Multiple Sclerosis were included in an open label, prospective study with a follow-up of 108 weeks. The primary endpoint (PE) was defined as “time to first relapse or MRI disease activity up to week 56”. In addition, a recurrent event analysis (REA) was performed up to week 108. RESULTSThe PE was not met (HR 0.67, 95% CI [0.22,1.97], p = 0.462). Number of relapses before stopping natalizumab was positively associated with the hazard of relapse (HR 3.91, p = 0.0117, 95% CI [1.36, 11.28]). The REA showed a reduction of the hazard to develop a relapse by 77% (HR 0.23, 95% CI [0.08, 0.69], p = 0.00854) in favor of the cohort with 4 weeks washout. CONCLUSIONSOur study suggests that switching from natalizumab to fingolimod with a shorter washout of 4 weeks might reduce the risk of disease reactivation after switching.

Factors associated with progression to pre-diabetes: a recurrent events analysis.

Pre-diabetes is a strong risk factor for type 2diabetes (T2D). The aim of this study was to explore factors associated with normal glucose maintenance and pre-diabetes prevention or delay. Data of 1016 first-degree relatives of T2D patients were retrieved from the Isfahan Diabetes Prevention Study (IDPS). Association of various variables including nutrients, serum tests and physical activity with the risk of pre-diabetes was assessed using recurrent events approach. Cumulative incidence of diabetes was 8.17, 9.44, and 4.91% for total sample and individuals with and without pre-diabetes experience in the follow-up. Risk of progression to pre-diabetes was higher in women and older people (p < 0.01). Additionally, BMI and blood pressure had significant association with the risk (p < 0.01) and individuals with higher intake of fat were at higher risk (HR = 2.26; 95% CI 1.66-3.07 for high-intake and HR = 1.52; 95% CI 1.27-1.83 for medium-intake compared to low-intake group). Carbohydrates and protein intake were positively associated with the risk of pre-diabetes with HR = 8.63 per 49 g extra carbohydrates per day and HR = 1.32 per 6 g extra protein per day (p < 0.01). The association was also significant for triglyceride (TG) with 7% risk increase per 1 SD = 1.14 increase in TG level. Despite frequent studies on lifestyle modification for pre-diabetes prevention, less information is available about the role of nutritional components. We observed direct effects for intake of macronutrients including fat, carbohydrates, and protein in first-degree relatives. Further research is warranted to assess these associations in general populations. Level III: Evidence obtained from a single-center cohort study.

Group sequential designs with robust semiparametric recurrent event models.

Robust semiparametric models for recurrent events have received increasing attention in the analysis of clinical trials in a variety of diseases including chronic heart failure. In comparison to parametric recurrent event models, robust semiparametric models are more flexible in that neither the baseline event rate nor the process inducing between-patient heterogeneity needs to be specified in terms of a specific parametric statistical model. However, implementing group sequential designs in the robust semiparametric model is complicated by the fact that the sequence of Wald statistics does not follow asymptotically the canonical joint distribution. In this manuscript, we propose two types of group sequential procedures for a robust semiparametric analysis of recurrent events. The first group sequential procedure is based on the asymptotic covariance of the sequence of Wald statistics and it guarantees asymptotic control of the type I error rate. The second procedure is based on the canonical joint distribution and does not guarantee asymptotic type I error rate control but is easy to implement and corresponds to the well-known standard approach for group sequential designs. Moreover, we describe how to determine the maximum information when planning a clinical trial with a group sequential design and a robust semiparametric analysis of recurrent events. We contrast the operating characteristics of the proposed group sequential procedures in a simulation study motivated by the ongoing phase 3 PARAGON-HF trial (ClinicalTrials.gov identifier: NCT01920711) in more than 4600 patients with chronic heart failure and a preserved ejection fraction. We found that both group sequential procedures have similar operating characteristics and that for some practically relevant scenarios, the group sequential procedure based on the canonical joint distribution has advantages with respect to the control of the type I error rate. The proposed method for calculating the maximum information results in appropriately powered trials for both procedures.

Clinical progression and metachronous paragangliomas in a large cohort of SDHD germline variant carriers.

Although it is well established that paternally transmitted germline variants in SDHD are associated with multifocal paragangliomas and lifelong follow-up is generally advised, the risk of metachronous lesions is presently unknown. In a large Dutch cohort of SDHD variant carriers, we studied the development of new paragangliomas, and the evolution of symptoms and cranial nerve impairment. Recurrent event analysis and the Kaplan-Meier product limit estimator were used to study the risk of new lesions. The relation between several predictors and development of new symptoms was assessed using logistic regression. Of the 222 SDHD variant carriers included, 65% presented with symptoms and 11% with cranial nerve dysfunction. Over a median period of 8 years, 42% reported new symptoms, and new cranial nerve impairment was observed in 11% of subjects. The estimated fraction of subjects that developed new HNPGL increased to 73% (95% CI: 52-85%) after 22 years of follow-up. Males were more likely to develop new HNPGL compared to females (HR: 1.63, 95% CI: 1.10-2.40), as were subjects that presented with symptoms, compared to subjects that were asymptomatic at baseline (HR: 1.61, 95% CI: 1.01-2.55). In addition, the risk of new lesions decreased with number of HNPGL present at first diagnosis (HR: 0.68 and 95% CI: 0.56-0.82). Carriers of a paternally inherited SDHD variant face a considerable risk for new HNPGL. In addition, nearly 50% of subjects reported new symptoms. However, new cranial nerve deficits were observed in only 11%, which is less than reported in surgical series. These risks should be taken into account when considering treatment strategies and counseling.

Stratified Cox Model with Interactions in Analysis of Recurrent Events

The purpose of this paper is the assessment of relative intensity of exit from registered unemployment by means of the analysis of recurrent survival episodes and the comparison of these results with the results obtained for an individual episode. The stratified Cox model with interactions was used. Statistical data collected by labour offices indicate that a large fraction of the unemployed persons is registered multiple times. However, many of them resign from the mediation of labour offices and are subsequently removed from the register. In the article, the intensities of de‑registration due to various causes for men and women were compared. The study data came from the database of personal details of people registered by the Poviat Labour Office in Szczecin in 2013. The observation covered the records of their registration until the end of 2014. Gender of the unemployed persons influenced the intensity of de‑registrations in the first episodes, partially in the second and third ones, due to various causes, such as finding a job or removal from the register, whereas it did not influence the intensity of de‑registrations in the fourth and subsequent episodes. As for the other causes in the subsequent episodes, the differences were also not statistically significant. The proposed analysis may be important for implementing a good policy in the labour market. The identification of persons that resign from the mediation of the labour office is as interesting as the identification of these persons who find a job.

The effect of omitted covariates in marginal and partially conditional recurrent event analyses

There have been many advances in statistical methodology for the analysis of recurrent event data in recent years. Multiplicative semiparametric rate-based models are widely used in clinical trials, as are more general partially conditional rate-based models involving event-based stratification. The partially conditional model provides protection against extra-Poisson variation as well as event-dependent censoring, but conditioning on outcomes post-randomization can induce confounding and compromise causal inference. The purpose of this article is to examine the consequences of model misspecification in semiparametric marginal and partially conditional rate-based analysis through omission of prognostic variables. We do so using estimating function theory and empirical studies.

Latent Class Analysis of Recurrent Events in Problem-Solving Items.

Computer-based assessment of complex problem-solving abilities is becoming more and more popular. In such an assessment, the entire problem-solving process of an examinee is recorded, providing detailed information about the individual, such as behavioral patterns, speed, and learning trajectory. The problem-solving processes are recorded in a computer log file which is a time-stamped documentation of events related to task completion. As opposed to cross-sectional response data from traditional tests, process data in log files are massive and irregularly structured, calling for effective exploratory data analysis methods. Motivated by a specific complex problem-solving item "Climate Control" in the 2012 Programme for International Student Assessment, the authors propose a latent class analysis approach to analyzing the events occurred in the problem-solving processes. The exploratory latent class analysis yields meaningful latent classes. Simulation studies are conducted to evaluate the proposed approach.

Adverse mental health outcomes in a population-based cohort of survivors of childhood cancer.

The elevated risk for physical late effects in childhood cancer survivors (CCS) is well documented, but their risk for mental health problems is less well described. The authors assembled a cohort of all 5-year CCS who were diagnosed before age 18 years and treated in an Ontario pediatric cancer center between 1987 and 2008. Patients were matched to population controls and linked to health administration databases. The authors calculated rates of mental health care visits (family physician, psychiatrist, emergency department, hospitalization) and the risk for a severe mental health event (emergency department, hospitalization, suicide). Outcomes were compared using recurrent event and survival analyses. Compared with 20,269 controls, 4117 CCS had a higher rate of mental health visits (adjusted relative rate [RR], 1.34; 95% confidence interval [CI], 1.12-1.52). Higher rates were associated with female gender (RR, 1.39; CI, 1.10-1.75; P = .006) and being diagnosed at ages 15 to 17.9 years (compared with ages 0-4 years: RR, 1.81; 95% CI, 1.17-2.80; P = .008). Cancer type, treatment intensity, and treatments targeting the central nervous system were not significant predictors. Survivors were at increased risk for a severe event compared with controls (adjusted hazard ratio, 1.13; 95% CI, 1.00-1.28; P = .045). CCS who were diagnosed with cancer at age 4 years or younger were at greatest risk: 16.3% (95% CI, 13.2%-19.8%) had experienced a severe event by age 28 years. CCS experienced higher rates of mental health visits and a greater risk for a severe event than the general population. Survivors of adolescent cancer have a higher rate of mental health visits overall, whereas survivors of cancer before age 4 years have a markedly elevated risk of severe events. Cancer 2018;124:2045-57. © 2018 American Cancer Society.

Time-to-first-event versus recurrent-event analysis: points to consider for selecting a meaningful analysis strategy in clinical trials with composite endpoints

Composite endpoints combining several event types of clinical interest often define the primary efficacy outcome in cardiologic trials. They are commonly evaluated as time-to-first-event, thereby following the recommendations of regulatory agencies. However, to assess the patient's full disease burden and to identify preventive factors or interventions, subsequent events following the first one should be considered as well. This is especially important in cohort studies and RCTs with a long follow-up leading to a higher number of observed events per patients. So far, there exist no recommendations which approach should be preferred. Recently, the Cardiovascular Round Table of the European Society of Cardiology indicated the need to investigate "how to interpret results if recurrent-event analysis results differ […] from time-to-first-event analysis" (Anker et al., Eur J Heart Fail 18:482-489, 2016). This work addresses this topic by means of a systematic simulation study. This paper compares two common analysis strategies for composite endpoints differing with respect to the incorporation of recurrent events for typical data scenarios motivated by a clinical trial. We show that the treatment effects estimated from a time-to-first-event analysis (Cox model) and a recurrent-event analysis (Andersen-Gill model) can systematically differ, particularly in cardiovascular trials. Moreover, we provide guidance on how to interpret these results and recommend points to consider for the choice of a meaningful analysis strategy. When planning trials with a composite endpoint, researchers, and regulatory agencies should be aware that the model choice affects the estimated treatment effect and its interpretation.

Effect of sacubitril/valsartan on recurrent events in the Prospective comparison of ARNI with ACEI to Determine Impact on Global Mortality and morbidity in Heart Failure trial (PARADIGM-HF).

AimsRecurrent hospitalizations are a major part of the disease burden in heart failure (HF), but conventional analyses consider only the first event. We compared the effect of sacubitril/valsartan vs. enalapril on recurrent events, incorporating all HF hospitalizations and cardiovascular (CV) deaths in PARADIGM‐HF, using a variety of statistical approaches advocated for this type of analysis.Methods and resultsIn PARADIGM‐HF, a total of 8399 patients were randomized and followed for a median of 27 months. We applied various recurrent event analyses, including a negative binomial model, the Wei, Lin and Weissfeld (WLW), and Lin, Wei, Ying and Yang (LWYY) methods, and a joint frailty model, all adjusted for treatment and region. Among a total of 3181 primary endpoint events (including 1251 CV deaths) during the trial, only 2031 (63.8%) were first events (836 CV deaths). Among a total of 1195 patients with at least one HF hospitalization, 410 (34%) had at least one further HF hospitalization. Sacubitril/valsartan compared with enalapril reduced the risk of recurrent HF hospitalization using the negative binomial model [rate ratio (RR) 0.77, 95% confidence interval (CI) 0.67–0.89], the WLW method [hazard ratio (HR) 0.79, 95% CI 0.71–0.89], the LWYY method (RR 0.78, 95% CI 0.68–0.90), and the joint frailty model (HR 0.75, 95% CI 0.66–0.86) (all P < 0.001). The effect of sacubitril/valsartan vs. enalapril on recurrent HF hospitalizations/CV death was similar.ConclusionsIn PARADIGM‐HF, approximately one third of patients with a primary endpoint (time‐to‐first) experienced a further event. Compared with enalapril, sacubitril/valsartan reduced both first and recurrent events. The treatment effect size was similar, regardless of the statistical approach applied.

Survival analysis of recurrent events on prostate cancer: facts from cancer genome

Many diseases and clinical outcomes may recur to the same patient. These events are termed as recurrent events. Several statistical models have been proposed in the literature to analyze recurrent events. In this study, we identify the clinical and the genetic risk factors for recurring tumors among prostate cancer patients from The Cancer Genome Atlas (TCGA). Five statistical approaches for modeling recurrent time-to-event are implemented to identify and to determine the effects of the clinical and the genetic risk factors of tumor recurrence. In particular, we consider Andersen-Gill (A-G), Wei-Lin-Weissfeld (WLW), Prentice-Williams-Peterson Total Time (PWP-TT), Prentice-Williams-Peterson Gap Time (PWPGT) and Frailty models. We present and discuss the risk factors influencing the recurrence of tumors and their impacts in prostate cancer patients obtained from five commonly used models in this paper.

The association between previous and future severe exacerbations of chronic obstructive pulmonary disease: Updating the literature using robust statistical methodology.

BackgroundThere is minimal evidence on the extent to which the occurrence of a severe acute exacerbation of COPD that results in hospitalization affects the subsequent disease course. Previous studies on this topic did not generate causally-interpretable estimates. Our aim was to use corrected methodology to update previously reported estimates of the associations between previous and future exacerbations in these patients.MethodsUsing administrative health data in British Columbia, Canada (1997–2012), we constructed a cohort of patients with at least one severe exacerbation, defined as an episode of inpatient care with the main diagnosis of COPD based on international classification of diseases (ICD) codes. We applied a random-effects 'joint frailty' survival model that is particularly developed for the analysis of recurrent events in the presence of competing risk of death and heterogeneity among individuals in their rate of events. Previous severe exacerbations entered the model as dummy-coded time-dependent covariates, and the model was adjusted for several observable patient and disease characteristics.Results35,994 individuals (mean age at baseline 73.7, 49.8% female, average follow-up 3.21 years) contributed 34,271 severe exacerbations during follow-up. The first event was associated with a hazard ratio (HR) of 1.75 (95%CI 1.69–1.82) for the risk of future severe exacerbations. This risk decreased to HR = 1.36 (95%CI 1.30–1.42) for the second event and to 1.18 (95%CI 1.12–1.25) for the third event. The first two severe exacerbations that occurred during follow-up were also significantly associated with increased risk of all-cause mortality. There was substantial heterogeneity in the individual-specific rate of severe exacerbations. Even after adjusting for observable characteristics, individuals in the 97.5th percentile of exacerbation rate had 5.6 times higher rate of severe exacerbations than those in the 2.5th percentile.ConclusionsUsing robust statistical methodology that controlled for heterogeneity in exacerbation rates among individuals, we demonstrated potential causal associations among past and future severe exacerbations, albeit the magnitude of association was noticeably lower than previously reported. The prevention of severe exacerbations has the potential to modify the disease trajectory.

An Overview of Statistical Models for Recurrent Events Analysis: A Review

An Overview of Statistical Models for Recurrent Events Analysis: A Review

Phase 3 Study of L-Glutamine in Sickle Cell Disease: Analyses of Time to First and Second Crisis and Average Cumulative Recurrent Events

▪BackgroundA Phase 3, double-blind, randomized, placebo-controlled trial of L-glutamine in 230 patients (ages 5-58) over 48 weeks was conducted at 31 centers in the United States. Twice daily oral administration of L-glutamine or placebo was based on body weight approximating 0.3 g/kg/dose. The primary endpoint was the number of sickle cell crises (SCCs) during 48 weeks of follow-up. The treatment effect measured was the difference of two median numbers of SCCs between two arms by week 48. The SCCs required treatment with I.V. narcotics/ketorolac at an emergency department or during hospitalization in patients with a history of ≥ 2 crisis from the previous year. Splenic sequestration, acute chest syndrome and priapism were also considered to be SCCs. Patients stabilized on hydroxyurea (HU) treatment for at least 3 months prior to study screening (66% in both arms of the study) remained on HU for the duration of the study. The primary analysis via a Cochran-Mantel-Haenszel (CMH) test statistic showed a highly significant treatment effect (p=0.005). There was a median of 3 crises for the L-glutamine treatment arm and 4 for the placebo arm (25% difference). There were no concerning adverse events from the treatment compared with the placebo arm. In this abstract, we reported the results from additional analyses of times to first and second crisis as well as cumulative recurrent events, which further demonstrated the clinically meaningful treatment effect of L-glutamine.MethodsThe Kaplan-Meier estimates are constructed with the data from the times to the first and second SCC for both arms. The corresponding estimates of the hazard ratios with confidence intervals and p-values are obtained to evaluate the relative merit of two groups. The 50% quartile (median) of the curve and the 95% confidence interval were reported by treatment. Furthermore, recurrent event time analysis was also performed using the Andersen-Gill procedure as well as the robust method by Lin, Wei, Yang and Ying to estimate the intensity rates.ResultsL-glutamine therapy significantly prolonged the time to first and the time to second SCCs. The median time to first crisis was delayed by 30 days (84 days vs. placebo 54 days; p=0.0152) and the median time to second crisis was delayed by 79 days (212 days vs. placebo 133 days; p=0.0260). Hazard ratios for the time to first and second crisis were similar (0.69 and 0.68 respectively), suggesting an approximately 30% hazard reduction from the treatment (Figures 1 and 2). Both cumulative recurrent event analyses models for SCCs yielded an intensity rate ratio of 0.75 with 95% confidence intervals based on the Andersen-Gill of (0.62, 0.90) and Lin, Wei, Yang and Ying of (0.55, 1.01), respectively (Figure 3). This suggested that the average cumulative crisis count was reduced by approximately 25% over the entire 48-week period. Cumulative event curve separation was clear by day 30 of treatment and was maintained over the 48-week study period.ConclusionThe results of the primary endpoint using time to event analytical methods substantiated the original CMH test result that L-glutamine treatment provided clinical benefit in reducing crises, and other acute complications of sickle cell disease, in patients 5 years of age and older irrespective of hydroxyurea use. EndariTM (L-glutamine oral powder) is now FDA-approved to reduce the acute complications of sickle cell disease in adult and pediatric patients 5 years of age and older. [Display omitted] DisclosuresNiihara:Emmaus Medical, Inc: Employment, Equity Ownership. Razon:Emmaus Medical, Inc: Employment. Wood:Emmaus Medical, Inc: Employment, Equity Ownership. Singh:Emmaus Medical, Inc: Employment. Tran:Emmaus Medical, Inc: Employment, Equity Ownership. Stark:Emmaus Medical, Inc: Employment, Equity Ownership.