Time-independent Covariates Research Articles

Delayed kernels for longitudinal survival analysis and dynamic prediction.

Predicting patient survival probabilities based on observed covariates is an important assessment in clinical practice. These patient-specific covariates are often measured over multiple follow-up appointments. It is then of interest to predict survival based on the history of these longitudinal measurements, and to update predictions as more observations become available. The standard approaches to these so-called 'dynamic prediction' assessments are joint models and landmark analysis. Joint models involve high-dimensional parameterizations, and their computational complexity often prohibits including multiple longitudinal covariates. Landmark analysis is simpler, but discards a proportion of the available data at each 'landmark time'. In this work, we propose a 'delayed kernel' approach to dynamic prediction that sits somewhere in between the two standard methods in terms of complexity. By conditioning hazard rates directly on the covariate measurements over the observation time frame, we define a model that takes into account the full history of covariate measurements but is more practical and parsimonious than joint modelling. Time-dependent association kernels describe the impact of covariate changes at earlier times on the patient's hazard rate at later times. Under the constraints that our model (a) reduces to the standard Cox model for time-independent covariates, and (b) contains the instantaneous Cox model as a special case, we derive two natural kernel parameterizations. Upon application to three clinical data sets, we find that the predictive accuracy of the delayed kernel approach is comparable to that of the two existing standard methods.

Blood pressure during long-term cilostazol-based dual antiplatelet therapy after stroke: a post hoc analysis of the CSPS.com trial

We determined the associations of follow-up blood pressure (BP) after stroke as a time-dependent covariate with the risk of subsequent ischemic stroke, as well as those of BP levels with the difference in the impact of long-term clopidogrel or aspirin monotherapy versus additional cilostazol medication on secondary stroke prevention. In a sub-analysis of a randomized controlled trial (CSPS.com), patients between 8 and 180 days after stroke onset were randomly assigned to receive aspirin or clopidogrel alone, or a combination of cilostazol with aspirin or clopidogrel. The percent changes, differences, and raw values of follow-up BP were examined. The primary efficacy outcome was the first recurrence of ischemic stroke. In a total of 1657 patients (69.5 ± 9.3 years, female 29.1%) with median 1.5-year follow-up, ischemic stroke recurred in 74 patients. The adjusted hazard ratio for ischemic stroke of a 10% systolic BP (SBP) increase from baseline was 1.19 (95% CI 1.03–1.36), that of a 10 mmHg SBP increase was 1.14 (1.03–1.28), and that of SBP as the raw value with the baseline SBP as a fixed (time-independent) covariate was 1.14 (1.00–1.31). Such significant associations were not observed in diastolic BP-derived variables. The estimated adjusted hazard ratio curves for the outcome showed the benefit of dual therapy over a wide SBP range between ≈120 and ≈165 mmHg uniformly. Lower long-term SBP levels after ischemic stroke were associated with a lower risk of subsequent ischemic events. The efficacy of dual antiplatelet therapy including cilostazol for secondary stroke prevention was evident over a wide SBP range.

Association of Dynamic Trajectories of Time-Series Data and Life-Threatening Mass Effect in Large Middle Cerebral Artery Stroke.

Life-threatening, space-occupying mass effect due to cerebral edema and/or hemorrhagic transformation is an early complication of patients with middle cerebral artery stroke. Little is known about longitudinal trajectories of laboratory and vital signs leading up to radiographic and clinical deterioration related to this mass effect. We curated a retrospective data set of 635 patients with large middle cerebral artery stroke totaling 95,463 data points for 10 longitudinal covariates and 40 time-independent covariates. We assessed trajectories of the 10 longitudinal variables during the 72h preceding three outcomes representative of life-threatening mass effect: midline shift ≥ 5mm, pineal gland shift (PGS) > 4mm, and decompressive hemicraniectomy (DHC). We used a "backward-looking" trajectory approach. Patients were aligned based on outcome occurrence time and the trajectory of each variable was assessed before that outcome by accounting for cases and noncases, adjusting for confounders. We evaluated longitudinal trajectories with Cox proportional time-dependent regression. Of 635 patients, 49.0% were female, and the mean age was 69years. Thirty five percent of patients had midline shift ≥ 5mm, 24.3% of patients had PGS > 4mm, and 10.7% of patients underwent DHC. Backward-looking trajectories showed mild increases in white blood cell count (10-11K/UL within 72h), temperature (up to half a degree within 24h), and sodium levels (1-3mEq/L within 24h) before the three outcomes of interest. We also observed a decrease in heart rate (75-65 beats per minute) 24h before DHC. We found a significant association between increased white blood cell count with PGS > 4mm (hazard ratio 1.05, p value 0.007). Longitudinal profiling adjusted for confounders demonstrated that white blood cell count, temperature, and sodium levels appear to increase before radiographic and clinical indicators of space-occupying mass effect. These findings will inform the development of multivariable dynamic risk models to aid prediction of life-threatening, space-occupying mass effect.

Evaluating the Performance of Mixed Zero-Inflated Poisson Regression Models with Time-dependent and Time-independent Covariates

One of the issues that researchers may encounter in count data is having many zeros. One of the solutions to model these data is using zero-inflated Poisson (ZIP) regression models. Recently, researchers have started to model longitudinal count data with time-dependent covariates. However, it has not been considered whether a model with time-dependent covariates provides a better fit than a model with time-independent covariates. In this paper, the fit between a mixed ZIP model with time-dependent covariates and a mixed ZIP model with time-independent covariates is compared using simulation. Using the deviance information criterion as a measure of fit, we found that the model with time-dependent covariates exhibits a better fit than the model with time-independent covariates. KEYWORDS correlated data, count data, excess zeros, longitudinal, mixed models, model fit

Covariate-guided Bayesian mixture of spline experts for the analysis of multivariate high-density longitudinal data.

With rapid development of techniques to measure brain activity and structure, statistical methods for analyzing modern brain-imaging data play an important role in the advancement of science. Imaging data that measure brain function are usually multivariate high-density longitudinal data and are heterogeneous across both imaging sources and subjects, which lead to various statistical and computational challenges. In this article, we propose a group-based method to cluster a collection of multivariate high-density longitudinal data via a Bayesian mixture of smoothing splines. Our method assumes each multivariate high-density longitudinal trajectory is a mixture of multiple components with different mixing weights. Time-independent covariates are assumed to be associated with the mixture components and are incorporated via logistic weights of a mixture-of-experts model. We formulate this approach under a fully Bayesian framework using Gibbs sampling where the number of components is selected based on a deviance information criterion. The proposed method is compared to existing methods via simulation studies and is applied to a study on functional near-infrared spectroscopy, which aims to understand infant emotional reactivity and recovery from stress. The results reveal distinct patterns of brain activity, as well as associations between these patterns and selected covariates.

Modeling of Distance Learners’ Retention Using Mixed-Model on Non-Proportional Hazard

Universitas Terbuka is a public university in Indonesia that implements an open-distance education system. The quality and success of higher education in Indonesia are indicated by the gross enrolment rate (GER). Modelling student retention is one indicator of the value of the GER. The chances of distance learners’ retention in a certain period can be determined through survival analysis, namely the “Cox model” or “Cox proportional hazard model”. The limitations of the Cox model pose new problems in modelling that involve more than two types of covariates and the presence of random effects. The problem that arises is the non-proportional hazard. The cause of non-proportional hazard is time-dependent covariates, so that individual risk changes over time. Another reason is the presence of random effects, both observed and unobserved. In this model, there are two random effects, namely, the unobserved random effect and the observed random effect. Thus, the model is called the mixed effect model of non-proportional hazard. Based on the characteristics of the covariates observed, appropriate survival analysis for modelling distance learners’ retention is a mixed model on the non-proportional hazard. The covariates involved in very complex modelling include time-independent covariates, time-dependent covariates, and unobserved random effects. The results of the analysis with the mixed model on non-proportional hazard showed that the covariates that significantly affected the distance learner’s retention were: educational background, age, grade point average, marital status, scholarship, independent learning culture, number of credits taken, and number of courses taken each semester.

Exploring the potential harm of varied blood storage on patients undergoing cardiovascular surgery.

Debate exists surrounding the optimal duration of red blood cell (RBC) storage. A hypothesis emerging from previous research suggests that exposure to fresh blood may be harmful to patients undergoing cardiac surgery. This study uses a large transfusion medicine database to explore the association between in-hospital mortality and red cell storage duration. This is an exploratory retrospective cohort study of all adult patients at Hamilton, Canada, over a 14-year period that received at least one allogeneic red cell transfusion during their hospitalization for cardiac surgery requiring bypass. The primary outcome for the study was in-hospital death. Analysis was performed using multivariate Cox regression modelling with time-dependent and time-independent covariates and stratification variables. Five models with varying definitions for short, intermediate and prolonged duration of RBC storage were tested. From March 2004 to December 2017, 11,205 patients met the inclusion criteria and were included in the regression analyses. No significant effect of short-duration red storage on patient mortality was observed in all statistical models, with the red cells stored for the longest duration as the reference group. When patients who received exclusively fresh (hazard ratio [HR] 1.040, 95% confidence interval [CI] 0.588-1.841, p-value = 0.893) and older aged (HR 1.038, 95% CI 0.769-1.1.402, p-value = 0.0801) RBCs were compared with those who received exclusively mid-age red cells as the reference, statistical significance was similarly not reached. Red cells stored for the shortest duration are not associated with increased risk of mortality among cardiac surgery patients.

Unleashing the Power of Very Small Data to Predict Acute Exacerbations of Chronic Obstructive Pulmonary Disease.

In this article, we explore to what extent it is possible to leverage on very small data to build machine learning (ML) models that predict acute exacerbations of chronic obstructive pulmonary disease (AECOPD). We build ML models using the small data collected during the eHealth Diary telemonitoring study between 2013 and 2017 in Sweden. This data refers to a group of multimorbid patients, namely 18 patients with chronic obstructive pulmonary disease (COPD) as the major reason behind previous hospitalisations. The telemonitoring was supervised by a specialised hospital-based home care (HBHC) unit, which also was responsible for the medical actions needed. We implement two different ML approaches, one based on time-dependent covariates and the other one based on time-independent covariates. We compare the first approach with standard COX Proportional Hazards (CPH). For the second one, we use different proportions of synthetic data to build models and then evaluate the best model against authentic data. To the best of our knowledge, the present ML study shows for the first time that the most important variable for an increased risk of future AECOPDs is "maintenance medication changes by HBHC". This finding is clinically relevant since a sub-optimal maintenance treatment, requiring medication changes, puts the patient in risk for future AECOPDs. The experiments return useful insights about the use of small data for ML.

Mixed Models of Non-Proportional Hazard and Application in The Open Distance Education Students Retention Data

The problem that arises in the Cox model is that there are more than two types of covariates and the presence of random effects is a non-proportional hazard (NPH). One example of a case that involves many factors is student retention. Low student retention can lead to dropping out of college or failure in completing studies. The purpose of this study is to overcome the problem of NPH caused by the presenceof time-independent covariates, time-dependent covariates, and random effects. The research method uses simulation. Some of the modified models are the stratified Cox model, the extended Cox model, and the frailty model. The developed model is applied to distance education student retention data. The results of the study show that frailty and study programs provide considerable diversity in explaining thetotal diversity of the model. It can be concluded that frailty needs to be considered by UT to improve the quality of services to students. In addition, other covariates that have a significant effect on UT student learning retention modeling are age, domicile, gender, GPA, marital status, employment status, number of credits taken, and number of registered courses.

Abstract 14609: Exposure to Any Hospitalization and Specific Invasive Procedures Post Open Heart Valve Surgery Increases the Risk of Endocarditis

Introduction: Infective endocarditis (IE) after cardiac valve surgery is associated with high morbidity and mortality. Nosocomial exposure is a growing cause of IE in general. We investigated the risk of post valve surgery endocarditis (PVE) in patient who had any hospitalization and specific nosocomial exposures after open heart valve surgery. Methods: We identified all ≥18yo patients who had their first open heart cardiac valve surgery between 2001-2017 in New South Wales, Australia from the Admitted Patient Data Collection (APDC) registry. Patients with prior/current IE diagnosis at time of index valve surgery were excluded. Follow up was until 31 Dec 2018 with mortality and morbidity tracked from the respective death and APDC registries. Analyses based on Cox regression modelling included age, sex, background diagnoses and features of index valve surgery as time independent covariates, with any hospitalization (separately for specific invasive procedures) post index valve surgery as time dependent exposure covariates for risk of PVE within 6 months of exposure. Results: In total 23747 patients (median age [IQR] 73yo [65-79yo, 63% male) had cardiac valve surgery: 60% isolated aortic valve (n=15065), 28% isolated mitral valve (n=6702), 10% multiple valves (n=2385) and 1.5% right sided valves (n=357). 5.4% (n=1293) of patients experienced PVE at a median 2.9 years (IQR 0.7-6.2) after index valve surgery, with 65% (n=838) occurring within 6 months of any hospitalization. Any hospitalization exposure was associated with an adjusted hazard ratio of 4.0 for developing PVE (95% confidence interval 3.5-4.5). In addition, specific invasive procedures including repeat valve surgery were associated with an elevated risk for PVE (Figure). Conclusions: PVE is significantly more common after any hospitalization with specific invasive procedures carrying differential risk. Care should be taken to avoid unnecessary hospitalizations and procedures in these patients.

Bayesian and Frequentist Approaches for a Tractable Parametric General Class of Hazard-Based Regression Models: An Application to Oncology Data

In this study, we consider a general, flexible, parametric hazard-based regression model for censored lifetime data with covariates and term it the “general hazard (GH)” regression model. Some well-known models, such as the accelerated failure time (AFT), and the proportional hazard (PH) models, as well as the accelerated hazard (AH) model accounting for crossed survival curves, are sub-classes of this general hazard model. In the proposed class of hazard-based regression models, a covariate’s effect is identified as having two distinct components, namely a relative hazard ratio and a time-scale change on hazard progression. The new approach is more adaptive to modelling lifetime data and could give more accurate survival forecasts. The nested structure that includes the AFT, AH, and PH models in the general hazard model may offer a numerical tool for identifying which of them is most appropriate for a certain dataset. In this study, we propose a method for applying these various parametric hazard-based regression models that is based on a tractable parametric distribution for the baseline hazard, known as the generalized log-logistic (GLL) distribution. This distribution is closed under all the PH, AH, and AFT frameworks and can incorporate all of the basic hazard rate shapes of interest in practice, such as decreasing, constant, increasing, V-shaped, unimodal, and J-shaped hazard rates. The Bayesian and frequentist approaches were used to estimate the model parameters. Comprehensive simulation studies were used to evaluate the performance of the proposed model’s estimators and its nested structure. A right-censored cancer dataset is used to illustrate the application of the proposed approach. The proposed model performs well on both real and simulation datasets, demonstrating the importance of developing a flexible parametric general class of hazard-based regression models with both time-independent and time-dependent covariates for evaluating the hazard function and hazard ratio over time.

Inference of a time-varying coefficient regression model for multivariate panel count data

Panel count data consist of the numbers of event occurrences between two consecutive observation times and are prevalent in many areas. Correspondingly a bulk of literature has been developed for the analysis of panel count data with time-independent or time-dependent covariates but assuming time-invariant covariate effects. However, the time-invariant coefficient assumption is too restrictive in reality and fails to represent the time-dynamic association between the covariates and event occurrence rates. In this paper, we discuss regression analysis of multivariate panel count data where both the covariates and their effects may be time-varying. We propose a marginal estimating equation approach combined with the B-splines that approximate the functional forms of the regression coefficients. The asymptotic properties of the proposed estimators are rigorously established. A simulation study is conducted to assess the performance of the proposed estimation procedure and suggests that it works well for practical situations. The proposed methodology is applied to a real dataset that motivated this study.

Using the Extended Cox Model to Determine Factors Affecting the Length of Hospitalization in Patients with Drug Poisoning

Background: Poisoning is a medical emergency, and is considered as a common cause of morbidity and mortality worldwide. In this study, the extended Cox model was used to determine the factors affecting the length of hospitalization in those with drug poisoning. Methods: The sample size included 2408 patients with opioids poisoning referring to the Emergency Department of Imam Reza Hospital in Mashhad, Iran from March 21, 2018 to March 20, 2019. Extended Cox model was fitted to determine the effect of five covariates (age, gender, marital status, type of poisoning, and type of opioids). In survival analysis, the length of hospitalization was considered as a time covariate (T). Patients’ recovery was also regarded as an event. Results: Of 2408 patients, 399 (16.6%) were censored and 2009 (83.4%) were uncensored. The risk of failure in complete recovery from poisoning in males was 1.189 times more compared to females. The risk of failure in complete recovery for the 15-24, 25-44, 45-64, and >65 years age groups were 0.277, 0.241, 0.289, and 0.481 times lower, respectively compared to the <2 years age group. For the married patients, the risk was 0.291 times lower compared to the divorced patients. For those poisoned accidentally, the risk was 0.490 times lower than compared to those poisoned intentionally. For those used methadone, morphine, opium, and tramadol, the risk was 1.195, 1.243, 1.193, and 1.147 times more, respectively compared to those used marijuana. By increasing the time (day) of hospital stay, the risk of failure for the 25-44, 45-64, and >65 years age groups were 1.024, 1.028, and 1.040 times more, respectively compared to the <2 years age group. Moreover, for those poisoned accidentally, the risk was 1.197 times more compared to those poisoned intentionally by the time (day) of hospital stay. Conclusion: The factors affecting the length of hospitalization in those poisoned by drugs are gender, marital status, and type of opioids covariate as time-independent covariate, and age and type of poisoning as time-dependent covariates. Since the complications of drug poisoning impose many costs on the health system, knowledge of these covariates can help take some measures for complete recovery of poisoned patients in a shorter length of hospital stay.

Using the Extended Cox Model to Determine Factors Affecting the Length of Hospitalization in Patients with Drug Poisoning

Background: Poisoning is a medical emergency, and is considered as a common cause of morbidity and mortality worldwide. In this study, the extended Cox model was used to determine the factors affecting the length of hospitalization in those with drug poisoning. Methods: The sample size included 2408 patients with opioids poisoning referring to the Emergency Department of Imam Reza Hospital in Mashhad, Iran from March 21, 2018 to March 20, 2019. Extended Cox model was fitted to determine the effect of five covariates (age, gender, marital status, type of poisoning, and type of opioids). In survival analysis, the length of hospitalization was considered as a time covariate (T). Patients’ recovery was also regarded as an event. Results: Of 2408 patients, 399 (16.6%) were censored and 2009 (83.4%) were uncensored. The risk of failure in complete recovery from poisoning in males was 1.189 times more compared to females. The risk of failure in complete recovery for the 15-24, 25-44, 45-64, and >65 years age groups were 0.277, 0.241, 0.289, and 0.481 times lower, respectively compared to the <2 years age group. For the married patients, the risk was 0.291 times lower compared to the divorced patients. For those poisoned accidentally, the risk was 0.490 times lower than compared to those poisoned intentionally. For those used methadone, morphine, opium, and tramadol, the risk was 1.195, 1.243, 1.193, and 1.147 times more, respectively compared to those used marijuana. By increasing the time (day) of hospital stay, the risk of failure for the 25-44, 45-64, and >65 years age groups were 1.024, 1.028, and 1.040 times more, respectively compared to the <2 years age group. Moreover, for those poisoned accidentally, the risk was 1.197 times more compared to those poisoned intentionally by the time (day) of hospital stay. Conclusion: The factors affecting the length of hospitalization in those poisoned by drugs are gender, marital status, and type of opioids covariate as time-independent covariate, and age and type of poisoning as time-dependent covariates. Since the complications of drug poisoning impose many costs on the health system, knowledge of these covariates can help take some measures for complete recovery of poisoned patients in a shorter length of hospital stay.

Generalized exponential distribution with interval-censored data and time dependent covariate

This study will improve the performance of the Generalized Exponential Distribution (GED) by incorporating time-dependent covariates (TD) in the presence of interval-censored data. Interval-censored data usually arises in clinical and epidemiological studies where lifetime is only known to fall within an interval. Moreover, this study concentrates on the parameters estimation for this distribution. This study will compare the maximum likelihood estimation (MLE) for two distinct models, namely time-dependent covariates model and time-independent covariates model in terms of their bias, standard error (SE) and root mean square error (RMSE) at various attendance probability (AP) and sample sizes. The results indicate that bias, SE and RMSE values of the parameter estimates increase with the increase in attendance probability and decrease with the increase in sample size.

Modeling Correlated Binary Outcomes with Time-Dependent Covariates

We group approaches to modeling correlated binary data accord ing to data recorded cross-sectionally as opposed to data recorded longi tudinally; according to models that are population-averaged as opposed to subject-specific; and according to data with time-dependent covariates as opposed to time-independent covariates. Standard logistic regression mod els are appropriate for cross-sectional data. However, for longitudinal data, methods such as generalized estimating equations (GEE) and generalized method of moments (GMM) are commonly used to fit population-averaged models, while random-effects models such as generalized linear mixed mod els (GLMM) are used to fit subject-specific models. Some of these methods account for time-dependence in covariates while others do not. This paper addressed these approaches with an illustration using a Medicare dataset as it relates to rehospitalization. In particular, we compared results from standard logistic models, GEE models, GMM models, and random-effects models by analyzing a binary outcome for four successive hospitalizations. We found that these procedures address differently the correlation among responses and the feedback from response to covariate. We found marginal GMM logistic regression models to be more appropriate when covariates are classified as time-dependent in comparison to GEE models. We also found conditional random-intercept models with time-dependent covariates decom posed into components to be more appropriate when time-dependent covari ates are present in comparison to ordinary random-effects models. We used the SAS procedures GLIMMIX, NLMIXED, IML, GENMOD, and LOGIS TIC to analyze the illustrative dataset, as well as unique programs written using the R language.

Functional Varying Coefficient Model with Time-independent Covariate and Longitudinal Response

In this paper, we consider functional varying coefficient model in present of a time invariant covariate for sparse longitudinal data contaminated with some measurement errors. We propose a regularization method to estimate the slope function based on a reproducing kernel Hilbert space approach. As we will see, our procedure is easy to implement. Our simulation results show that the procedure performs well, especially when either sampling frequency or sample size increases. Applications of our method are illustrated in an analysis of a longitudinal CD4+ count dataset from an HIV study.

Regression Estimation for Longitudinal Data with Nonignorable Intermittent Nonresponse and Dropout

We mainly focus on regression estimation in a longitudinal study with nonignorable intermittent nonresponse and dropout. To handle the identifiability issue, we take a time-independent covariate as nonresponse instrument which is independent of nonresponse propensity conditioned on other covariates and responses to ensure the identifiability of nonresponse propensity. The nonresponse propensity is assumed to be a parametric model, and the corresponding parameters are estimated by using the generalized method of moments approach. Then the marginal response means are estimated by inverse probability weighting method. Furthermore, to improve the robustness of estimators, we derive an augmented inverse probability weighting estimator which is shown to be consistent and asymptotically normally distributed. Simulation studies and a real-data analysis show that the proposed approach yields highly efficient estimators.

Determining the effect of time dependent and time independent factors on pneumonia of children under five in North west Ethiopia.

Lower Respiratory Tract Infections including pneumonia are the fourth cause of death globally. In Ethiopia, pneumonia is the leading cause of death for children under five. However, so far, only a few studies that used longitudinal design and time dependent covariates determined the significant factors of pneumonia. This study sought to determine whether respiratory rate changes differed for under five-year-old pneumonia patients who had been receiving different treatments over time and whether the change was effected by time dependent and independent covariates. A longitudinal study design involving marginal Poisson regression models and conditional Poisson regression models was used. After comparing the two, the final interpretation was made using a conditional Poisson regression model owing to its relative powerfulness. Four hundred and fifty-three under five pneumonia patients were included, of which 44.37% were female. It is found that, compared to rural children, urban children had an estimated mean respiratory rate decrease of 3%. It is also found that, compared to children whose mother practiced only exclusive breastfeeding in the first six months, children whose mother practiced both breastfed and complementary were more likely to be exposed to pneumonia. The estimated mean respiratory rate of children having asthma was 1.073 times that of children who had diarrhea. In northwest Ethiopia, weight, residence, previous disease history, breastfeeding and temperature are significant factors of pneumonia among children under five. The effectiveness of treatments was dependent on the number of times children visited the hospital. A significant variation of baseline pneumonia status among under five pneumonia patient children was noted in the hospital.

New onset diabetes predicts progression of low risk pancreatic mucinous cysts

BackgroundPatients with low-risk lesions require ongoing surveillance since the rate of progression to pancreatic cancer (PC), while small, is much greater than in the general population. Our objective was to study the relationship between new onset diabetes (NODM) and progression in patients with low risk mucinous cysts. MethodsWe evaluated a prospectively maintained cohort of 442 patients with a suspected mucinous cyst without worrisome features (WF) or high-risk stigmata (HRS). Multivariable Cox models were developed for progression to WF and HRS, with diabetes status formulated as both time independent and dependent covariates. The adjusted cumulative risk of progression was calculated using the corrected group prognosis method. ResultsThe 5-year cumulative progression rates to WFs and HRS were 12.8 and 3.6%, respectively. After controlling for other risk factors, the development of NODM was strongly associated with progression to HRS (HR = 11.6; 95%CI, 3.5–57.7%), but not WF. Among patients with the smallest cysts (<10 mm) at baseline, those who developed NODM had a 5-year adjusted cumulative risk of progression to HRS of 8.6% (95%CI, 0.0%–20.2%), compared to only 0.8% (95%CI, 0.0%–2.3%) for patients without NODM. Among patients with the largest cysts (20–29 mm), those who developed NODM during surveillance had a 5-year adjusted cumulative risk of progression of 53.5% (95%CI, 19.6%–89.9%) compared to only 7.5% (95%CI, 1.6%–15.2%) for patients without NODM. ConclusionNew onset diabetes may predict progression in patients with low risk mucinous cysts. Pending validation with large-scale studies, these findings support regular diabetes screening among patients surveilled for suspected IPMNs or MCNs.