Survival Analysis Models Research Articles

Survival Analysis of Immunotherapy Effects on Relapse Rate in Pediatric and Adult Autoimmune Encephalitis.

Prior observational studies for autoimmune encephalitis (AE) have mostly focused on outcomes after acute immunotherapies with better outcomes associated with earlier immunotherapy use. However, the impact of long-term immunotherapy and its association with clinical relapse is not well known. We conducted a retrospective study of consecutive patients meeting published clinical criteria for AE evaluated at UC San Diego and Rady Children's Hospital from January 2007 to November 2021. Survival analysis and Cox multivariable regression models were used to evaluate relapse risk using rituximab exposure as a time-dependent variable. Pooled and age-stratified analyses were performed. A total of 204 pediatric and 380 adult participants were screened of which 30 pediatric and 75 adult participants were included. The most common antibody subtype in both cohorts was anti-NMDA receptor (76% in pediatric, 34% in adult). Relapses occurred in 31% of pediatric antibody-positive, 40% of adult antibody-positive, and 20% of adult antibody-negative cases. Times to first relapse (TTFR) were 10.6 ± 7.4 months (pediatric antibody-positive), 13.1 ± 24.5 months (adult antibody-positive), and 6.9 ± 3.8 months (adult antibody-negative). Rituximab was the most common second-line immunotherapy used. Combining pediatric and adult data, rituximab use was associated with a 71% lower hazard for time to first relapse (hazard ratio [HR] 0.29, 95% CI 0.09-0.85) and 51% lower hazard for recurring relapses (HR 0.49, 95% CI 0.9-1.26). The HR for TTFR with rituximab use in children was 0.30 (95% CI 0.05-1.69), 0.29 (95% CI 0.07-1.29) in adults, 0.32 in non-NMDA antibody-positive encephalitis (95% CI 0.07-1.39), and 0.42 (95% CI 0.07-2.67) for anti-NMDAR. Relapses are common in pediatric and adult patients with AE, although less frequently in anti-NMDARE. Using a rigorous survival model, we demonstrate a substantial benefit of rituximab use for reducing relapse rates in AE, especially for the adult population. This study provides Class IV evidence that rituximab is associated with a lower hazard to relapse in patients with AE.

Sensitivity of Survival Analysis Metrics

Survival analysis models allow for predicting the probability of an event over time. The specificity of the survival analysis data includes the distribution of events over time and the proportion of classes. Late events are often rare and do not correspond to the main distribution and strongly affect the quality of the models and quality assessment. In this paper, we identify four cases of excessive sensitivity of survival analysis metrics and propose methods to overcome them. To set the equality of observation impacts, we adjust the weights of events based on target time and censoring indicator. According to the sensitivity of metrics, AUPRC (area under Precision-Recall curve) is best suited for assessing the quality of survival models, and other metrics are used as loss functions. To evaluate the influence of the loss function, the Bagging model uses ones to select the size and hyperparameters of the ensemble. The experimental study included eight real medical datasets. The proposed modifications of IBS (Integrated Brier Score) improved the quality of Bagging compared to the classical loss functions. In addition, in seven out of eight datasets, the Bagging with new loss functions outperforms the existing models of the scikit-survival library.

PICBayes: Bayesian proportional hazards models for partly interval-censored data

Partly interval-censored (PIC) data arise frequently in medical studies of diseases that require periodic examinations for symptoms of interest, such as progression-free survival and relapse-free survival. The proportional hazards (PH) model is the most widely used model in survival analysis. This paper introduces our new R package, PICBayes, which implements a set of functions for fitting the PH model to different complexities of partly interval-censored data under the Bayesian semiparametric framework. The main function of PICBayes is to fit (1) the PH model to PIC data; (2) the PH model with spatial frailty to areally-referenced PIC data; (3) the PH model with one random intercept to clustered PIC data; (4) the PH model with one random intercept and one random effect to clustered PIC data; and (5) general mixed effects PH model to clustered PIC data. We also included the corresponding functions for general interval-censored data. A random intercept or random effect can follow both a normal prior and a Dirichlet process mixture prior. The use of the package is illustrated by analyzing two real data sets.

Progress in transforming a health sciences postgraduate cohort in a south african research-intensive institution, 2008–2017

BackgroundEquity redress in the higher education and health sectors is a global discourse that seeks to address the inequalities caused by past discrimination practices. The apartheid regime in South Africa fragmented both the higher education and the health sectors, creating White and male dominated systems. Consequently, Black Africans and females were under-represented in these sectors. Furthermore, the provision of higher education including medical training was unequal between the different populations. As democracy was established in South Africa in 1994, it is necessary to assess whether transformation in population affinity and sex of postgraduate students in the higher education and health sector has occurred, as these individuals are crucial for providing the future academic workforce and also healthcare to the public.MethodsThe demographic profile of postgraduate students graduating in a health sciences facility in South Africa over the period 2008–2017 was retrospectively assessed. Survival analysis models were used to investigate the time taken to graduate. Log-rank tests were used to compare the completion rates.ResultsMore females (53.3%) than males (41.9%) completed their postgraduate degree over the period 2008–2017 (p˂0.0001). In relation to population affinity, more White students (56.4%) than Black African students (40.8%) completed their degrees overall (p˂0.0001).ConclusionWhile transformation occurred in the sex of graduating students over the ten year period, the same change has not occurred with regards to population affinity. The under-representation of Black African graduates is a major setback for efforts to diversify the South African higher education and health sectors. Transformation of the demographic profile of postgraduate students at South African institutions is vital for developing individuals who will contribute to equitable redress of academic staff in the higher education sector and also of the healthcare workforce. Diversified health personnel including highly skilled clinician scientists will aid in improving the provision of health care to communities particularly the underpriviledged rural areas, and also assist in training the next generation of healthcare staff. The challenges identified in this study may assist other countries where adequate transformation of the education and health sectors has not occurred.

Guiding the postoperative radioactive iodine-131 therapy for patients with papillary thyroid carcinoma according to the prognostic risk groups: a SEER-based study.

The effectiveness of iodine-131(131I) therapy in patients with papillary thyroid cancer (PTC) of various stage is controversial. This study aimed to use prognostic risk groups to guide 131I therapy in patients with PTC after radical thyroidectomy. Data of 53,484 patients with PTC after radical thyroidectomy were collected from the Epidemiology and End Results (SEER) database. Patients were divided into subgroups according to MACIS system and regional lymph node involvement. The prognostic role of 131I therapy was investigated by comparing Kaplan-Meier survival analysis and Cox proportional hazard models in different subgroups. Sex, age, tumor size, invasion, regional lymph node involvement, and distant metastasis was related to the survival of patients with PTC. If MACIS < 7, 131I treatment didn't affect the cancer-specific survival (CSS) rate. If MACIS ≥ 7, 131I therapy didn't work on CSS rate for patients with N0 or N1a < 5 status; 131I therapy had improved CSS rate for patients in the N1a ≥ 5 or N1b status. If patients with distant metastasis, invasion, or large tumor, 131I therapy didn't improve CSS rate for patients in N0 or N1a < 5 stage. After radical thyroidectomy, if MACIS < 7, patients with PTC could avoid 131I therapy. If MACIS ≥ 7, patients in the N0 or N1a < 5 could avoid 131I therapy; those in the N1a ≥ 5 or N1b stage should be given 131I therapy. Among them, all patients with distant metastasis should be given 131I therapy.

On the Reliability of Machine Learning Models for Survival Analysis When Cure Is a Possibility

In classical survival analysis, it is assumed that all the individuals will experience the event of interest. However, if there is a proportion of subjects who will never experience the event, then a standard survival approach is not appropriate, and cure models should be considered instead. This paper deals with the problem of adapting a machine learning approach for classical survival analysis to a situation when cure (i.e., not suffering the event) is a possibility. Specifically, a brief review of cure models and recent machine learning methodologies is presented, and an adaptation of machine learning approaches to account for cured individuals is introduced. In order to validate the proposed methods, we present an extensive simulation study in which we compare the performance of the adapted machine learning algorithms with existing cure models. The results show the good behavior of the semiparametric or the nonparametric approaches, depending on the simulated scenario. The practical utility of the methodology is showcased through two real-world dataset illustrations. In the first one, the results show the gain of using the nonparametric mixture cure model approach. In the second example, the results show the poor performance of some machine learning methods for small sample sizes.

Prediction of 30-day unplanned hospital readmission through survival analysis

Background and ObjectiveUnplanned hospital readmissions are a severe and recurrent problem that affects all health systems. Estimating the risk of being readmitted the following days after discharge is difficult since many heterogeneous factors can influence this. The extensive work concerning this problem proposes solutions mostly based on classification machine-learning models. Survival analysis methods could make a better match with the assessment of readmission risk and are yet to become well-established in this field. MethodsWe compare different statistical and machine learning survival analysis models trained with right-censored all-cause hospital admission data with covariates available at the moment of discharge. The main focus is on tree-ensemble regression methods based on the assumption of proportional hazards. These models are more thoroughly evaluated at a 30-day time period after discharge, although the actual prediction could be set to any time up to 90 days. ResultsThe mean performance obtained by each of the proposed survival models ranges from 0.707 to 0.716 C-Index and 0.709 to 0.72 ROC-AUC at a 30-day time period after discharge. The model with the lower performance on both metrics was Cox Proportional Hazards, while the model marking the upper end on both ranges is an XGBoost Regression model with a Cox objective function. ConclusionsOur findings indicate that survival models perform well addressing the hospital readmission problem, machine-learning models getting the edge over statistical methods. There seems to be an improvement over classification models when attempting to predict at a 30-day period since discharge, perhaps due to a better handling of cases nearing the 30-day boundary. Some preprocessing steps, such as limiting the observation period to 90 days after discharge, are also highlighted since they resulted in a performance boost.

Incidence of Rifampicin Resistance in Periprosthetic Joint Infection: A Single-Centre Cohort Study on 238 Patients.

Rifampicin is a pillar in the treatment of periprosthetic joint infection (PJI). However, rifampicin resistance is an increasing threat to PJI treatment. This study explores the incidence of rifampicin-resistant bacteria over time in a Swedish tertiary referral centre and the association of rifampicin resistance with infection-free survival after PJI. The study included 238 staphylococcal PJIs treated between 2001 and 2020 for which susceptibility data for rifampicin were available. Data on causative bacteria, rifampicin resistance, treatment, and outcome were obtained. Kaplan-Meier survival analysis and Cox regression modelling estimated the infection-free cumulative survival and adjusted hazard ratios (HRs) for the risk of treatment failure. Rifampicin-resistant causative bacteria were identified in 40 cases (17%). The proportion of rifampicin-resistant agents decreased from 24% in 2010-2015 to 12% in 2016-2020. The 2-year infection-free survival rates were 78.6% (95% CI, 66.4-93.1%) for the rifampicin-resistant group and 90.0% (95% CI, 85.8-94.4%) for the rifampicin-sensitive group. Patients with PJI caused by rifampicin-resistant bacteria had an increased risk of treatment failure (adjusted HR, 4.2; 95% CI, 1.7-10.3). The incidence of PJI caused by rifampicin-resistant bacteria did not increase over the past 20 years. The risk of treatment failure in PJI caused by rifampicin-resistant bacteria is more than four times that caused by rifampicin-sensitive bacteria, highlighting the importance of limiting the development of rifampicin resistance.

Interspecific Repellent Effect of Hydroalcoholic Solution of Nest Refuse from Two Leaf-cutter Ant Species

Leaf-cutting ants manage their waste continuously to avoid risks of colony contamination. The presence of residues or their odor near foraging areas is enough for ants to stop attacking plant tissues. However, to date, no study has demonstrated an interspecific effect of nest waste on the deterrence of leaf-cutting ants. Here, we tested whether a hydroalcoholic extract of nest refuse (NR) of the leaf-cutting ants Acromyrmex balzani Emery, 1890 and Atta opaciceps Borgmeier (Hymenoptera: Formicidae, Attini) causes an interspecific deterrent effect. Field colonies of Ac. balzani (n = 30) and A. opaciceps (n = 36) were used. For each colony, we offered baits made with plastic straws covered by citrus pulp and sprayed with a hydroalcoholic solution of 20% (vol/vol) alcohol/water (Control) or an extract of 20% (vol/vol) of NR from Ac. balzani and/or A. opaciceps. Over 20-30 minutes we recorded the amount of baits removed and data were compared through survival analysis and Linear Mixed Effect (LME) model. Ants avoided collecting baits sprayed with NR resulting in less than 20% of these baits being carried to the nest versus 80% of control baits. The deterrence was both intra and interspecific, considering that the response was similar for baits from both species. Such results reinforce the premise that the presence of nest refuse alters the foraging behavior of ants and paves the way for new studies that can verify its effectiveness in reducing foliar consumption on a larger spatial scale.

Evaluation of Different Survival Analysis Models for Nki Breast Cancer Data

The goal of this study is to evaluate different Survival Analysis Models in terms of their predictive capabilities, accuracy in determining significant covariates within the data, as well as their respective results compared across standard indices. Highest Concordance Index and Lowest Akaike Information Criterion (AIC) are used as the basis of selecting the ideal Survival Analysis model as a template for the construction of the Survival Prediction model for NKI Breast Cancer Data. 6 Survival Analysis Models were used in this study. For the semi-parametric survival models, Classical Cox, Cox-Lasso, and Cox-Ridge Regressions. For the parametric models, 3 Accelerated Failure Time (AFT) models were implemented. These are: Weibull AFT, Log-logistic AFT, and Log-Normal AFT models. Right-censoring was performed in the data since it has been assumed that there are subjects which were not called back anymore for the entire, 18-year clinical trial where the data was taken from. A proportional hazards test was then performed to find out if the covariates in the data are fit to be modeled using Cox Regression and its derivatives. A test for the distribution on the time of event was also done to find whether it follows a specific distribution or not. This was done to verify the usability of the parametric survival analysis models on the data. It has been found out that in terms of Concordance Index and AIC, the Cox-Ridge Regression model outperforms its 2 other semi-parametric counterparts, having the least AIC of 752.6703 and Highest Concordance Index of 0.7709. As for the other 3 parametric models, Log-Normal AFT outperformed the Weibull AFT and Log-Logistic AFT models by a Concordance Index of 0.780 with a corresponding AIC of 608.822. This result also suggests that the time of event of the subjects is best fitted by Log-Normal Distribution. By comparing the 2, best-performing models, it has been reported that Log-Normal AFT outperforms Cox-Ridge Regressions, therefore suggesting to use this Parametric Survival Analysis Model as the basis for a Survival Prediction model suited for NKI Breast Cancer data.

Statistical methods and data visualisation of patient-reported outcomes in early phase dose-finding oncology trials: a methodological review

Traditionally, within dose-finding clinical trials, treatment toxicity and tolerability are assessed by clinicians. Research has shown that clinician reporting may have inadequate inter-rater reliability, poor correlation with patient reported outcomes, and under capture the true toxicity burden. The introduction of patient-reported outcomes (PROs), where the patient can assess their own symptomatic adverse events or quality of life, has potential to complement current practice to aid dose optimisation. There are no international recommendations offering guidance for the inclusion of PROs in dose-finding trial design and analysis. Our review aimed to identify and describe current statistical methods and data visualisation techniques employed to analyse and visualise PRO data in published early phase dose-finding oncology trials (DFOTs). DFOTs published from June 2016-December 2022, which presented PRO analysis methods, were included in this methodological review. We extracted 35 eligible papers indexed in PubMed. Study characteristics extracted included: PRO objectives, PRO measures, statistical analysis and visualisation techniques, and whether the PRO was involved in interim and final dose selection decisions. Most papers (30, 85.7%) did not include clear PRO objectives. 20 (57.1%) papers used inferential statistical techniques to analyse PROs, including survival analysis and mixed-effect models. One trial used PROs to classify a clinicians' assessed dose-limiting toxicities (DLTs). Three (8.6%) trials used PROs to confirm the tolerability of the recommended dose. 25 trial reports visually presented PRO data within a figure or table within their publication, of which 12 papers presented PRO score longitudinally. This review highlighted that the statistical methods and reporting of PRO analysis in DFOTs are often poorly described and inconsistent. Many trials had PRO objectives which were not clearly described, making it challenging to evaluate the appropriateness of the statistical techniques used. Drawing conclusions based on DFOTs which are not powered for PROs may be misleading. With no guidance and standardisation of analysis methods for PROs in early phase DFOTs, it is challenging to compare study findings across trials. Therefore, there is a crucial need to establish international guidance to enhance statistical methods and graphical presentation for PRO analysis in the dose-finding setting. EA has been supported to undertake this work as part of a PhD studentship from the Institute of Cancer Research within the MRC/NIHR Trials Methodology Research Partnership. AM is supported by the National Institute for Health Research (NIHR) Biomedical Research Centre at the Royal Marsden NHS Foundation Trust, the Institute of Cancer Research and Imperial College.

Association between diversity levels of member composition in group activities of older adults and the occurrence of need for care: the JAGES 2013–2019 longitudinal study

BackgroundParticipating in groups with diverse members is associated with improved health among older adults. The study examined the relationship between diversity of group members and needed support or long-term care.MethodsWe conducted a longitudinal study for the Japan Gerontological Evaluation Study with 61,281 participants aged ≥ 65 years who were surveyed in 2013 and followed-up for six years. We assessed three dimensions of the diversity of the participating members (sex, age, and region of residence). We then graded the diversity level into four categories: level 0 (not in any group), level 1 (in a group without diversity or in a group with diversity in one of the three factors), level 2 (in a group with diversity in two of the three factors), or level 3 (in a group with diversity across all factors). We adjusted for 12 covariates using Cox hazard survival analysis models with hazard ratios (HRs) and 95% confidence intervals (CIs) estimated for the association between group members’ diversity levels and needed support or long-term care. The same study was conducted when stratified by employment status at baseline.ResultsParticipants in social participation groups with more diverse group members had a lower incidence of needed support or long-term care as compared to their counterparts. Compared to those with no participation group, HR decreased by 14% to 24% with increasing levels of diversity. The HR for the level of care needed for participants in the social participation group with high residential diversity was 0.89 (95% CI: 0.84–0.94). For participants who were currently unemployed, HR reductions ranged from 16%–28% with increasing levels of diversity compared to the non-participating group. No association was found for employed participants.ConclusionsThe reason the HRs of Japanese elderly people certified as needing support or care are lower when the diversity of participating groups is higher could be owing to the presence of a variety of people and the diversification of social networks, which facilitates the building of bridging social relational capital. Public health policies that encourage participation in diverse organizations will be important in the future.

Cost-effectiveness analysis of nivolumab plus chemotherapy vs chemotherapy for patients with unresectable advanced or metastatic HER2-negative gastric or gastroesophageal junction or esophageal adenocarcinoma in Japan.

This study aimed to evaluate the cost-effectiveness of nivolumab plus chemotherapy (NIVO + Chemo) compared with chemotherapy monotherapy (Chemo) for patients with advanced or metastatic HER2-negative gastric or gastroesophageal junction or esophageal adenocarcinoma (GC/GEJC/EAC) in Japan from the perspective of healthcare payer. A partitioned survival analysis model was developed to predict costs and quality-adjusted life years (QALYs) for NIVO + Chemo and Chemo. The time horizon of the model was set to 38years. An annual discount rate of 2% for both costs and QALYs was applied. Data on overall survival and progression-free survival were derived from the CheckMate649 trial. Cost parameters were estimated from a Japanese medical claims database. The incremental cost-effectiveness ratio (ICER) of NIVO + Chemo compared with Chemo was estimated. A subgroup analysis on the level of PD-L1 CPS expression was conducted. In addition, sensitivity analysis was performed to assess the uncertainty in the parameter settings. The incremental cost and QALY of NIVO + Chemo compared with Chemo were USD99,416 and 0.30 QALY, respectively. The ICER of NIVO + Chemo was estimated to be USD327,161 per QALY gained. The results of the subgroup analysis showed that ICER was USD247,403/QALY and USD302,183/QALY for PD-L1 CPS ≧ 5 and ≧ 1, respectively. Sensitivity analyses showed a relatively robust result that the ICER remained higher than the Japanese cancer threshold of USD75,000-150,000/QALY. Applying the Japanese cancer threshold of USD75,000-150,000/QALY, NIVO + Chemo was not cost-effective for patients with advanced or metastatic HER2-negative GC/GEJC/EAC in Japan from the perspective of healthcare payer.

Long term respiratory morbidity of cesarean-delivered second twin compared to their vaginally-delivered sibling: A retrospective population-based cohort study.

Offspring born via cesarean delivery (CD) may be more prone to develope long-term respiratory diseases, compared to those delivered vaginally (VD). In this study, we compared the rates of respiratory diseases between first twins VDand second twins delivered via CD. This was a retrospective database study. All twin deliveries encompassed at the Soroka University Medical Center, a large tertiary hospital in southern Israel, between 1991 and 2020, in which the first twin was VDand the second via CD were included. Respiratory diseases included respiratory tract diseases such as bronchiolitis and bronchial asthma. The cumulative incidence of respiratory diseases was compared between the twins using Kaplan-Meier survival analysis and multivariable Cox models to adjust for confounding variables. A total of 395,408 deliveries occurred during the study period, with 13,402 (3.4%) of all deliveriesbeing twins. Of these, 184 (1.4%) were first twins VDand second twins delivered via CD. The second CD twin was more likely to have a non-reassuring fetal heart rate pattern and an Apgar score less than 7 at 5 min. No other differences were found between the siblings. The incidence of long-term respiratory diseases was not statistically different between the CD and VD siblings (7.6% vs. 9.4%, respectively; OR = 0.54; 95%CI: 0.23-1.26). Similarly, the cumulative incidence of respiratory diseases was not statistically different (Kaplan-Meier, log-rank, p = .59), and in the multivariable analysis which adjusted for birthweight and fetal distress during delivery (adjusted hazard ratio = 1.06; 95%CI: 0.43-26.25). While the immediate outcomes for the CD twin were slightly worse compared to the VD twin, there was no difference in long-term respiratory diseases between the siblings.

Reliability and forced outages: Survival analysis with recurrent events

This paper empirically investigates the contributory factors to forced outage declarations of power plants through a survival analysis model with recurrent events. Using plant-level data in the Philippines, we find that higher reserve margins, higher share capacity (a measure of concentration), and the number of planned outages and days since the last maintenance (preventive maintenance indicators) reduce the risk of forced outages. On the other hand, an increase in the use rate (a measure of the intensity of power plant utilization) and geothermal, solar, and biodiesel plants correlate positively to more forced outages. Focusing on episodes where outages were disproportionately higher in the past decade— we find that thinning reserve margin correlates significantly to forced outage incidence. We also find that a higher share of a power firm’s capacity to the system’s total dependable capacity contributes to lowering the risk of forced outages.

Cardiovascular events more than 6 months after pregnancy in patients with congenital heart disease

ObjectivesPatients with congenital heart disease (CHD) are increasingly pursuing pregnancy, highlighting the need for data on late cardiovascular events (more than 6 months after delivery). We aimed to determine the...

Bayesian and non-Bayesian estimations of truncated inverse power Lindley distribution under progressively type-II censored data with applications

In this article, we introduce and study the truncated inverse power Lindley distribution. The aim is to transpose the remarkable flexibility of the two-parameter inverse power Lindley distribution to the interval [0,1]. The corresponding probability density function has the potential to be unimodal, decreasing, right-skewed, and heavy-tailed. On the other hand, the hazard rate function can be increasing, N-shaped, or U-shaped. These shapes’ versatility enables accurate representation and analysis of proportional or percentage data across a wide range of applications, such as survival analysis, reliability, and uncertainty modeling. Several statistical features, such as the mode, quantiles, Bowley’s skewness, Moor’s kurtosis, MacGillivray’s skewness, moments, inverse moments, incomplete moments, and probability-weighted moments, are computed. In practice, for the estimation of the model parameters from truncated data under the progressively type-II censoring scheme, the maximum likelihood, maximum product spacing, and Bayesian approaches are used. The Tierney–Kadane approximation and Markov chain Monte Carlo techniques are employed to produce the Bayesian estimates under the squared error loss function. We present some simulation results to evaluate these approaches. Four applications based on real-world datasets—one of them is on times of infection, the second is on failure times, and the other two are on the rate of inflation in Asia and Africa—explain the significance of the new truncated model in comparison to some reputed comparable models, such as the inverse power Lindley, Kumaraswamy, truncated power Lomax, beta, truncated Weibull, unit-Weibull, Kumaraswamy Kumaraswamy, and exponentiated Kumaraswamy models.

Protective effect of early immunomodulatory treatment on ocular involvement in Behcet's disease: Historical cohort of 1166 patients

ObjectivesEye involvement is a main presentation of Behcet's disease. This study was performed to evaluate possible determinants affecting the occurrence of eye involvement, especially the role of early systemic treatment with immunomodulatory drugs on the incidence of ocular involvement. MethodsThis is a retrospective cohort study performed on 1166 Behcet's patients in the Behcet's Clinic of Rheumatology Research Center. All patients were followed up for at least 10 years and a maximum of 15 years. Data analysis was performed using survival analysis models including Kaplan-Meier Survival analysis, Logrank test, and Cox's proportional hazards regression. Results1166 Behcet's patients were evaluated. 80 patients who had eye involvement as the first manifestation of the disease were excluded and 1086 participants entered the analysis. Among them, 647 patients (59.6%) developed ocular involvement 7.8 ± 6.7 years after the first symptom. Immunomodulatory treatment before ocular involvement reduced the risk by 3 times (P-value <0.001). ConclusionThis study demonstrated that the initiation of immunomodulatory treatment prior to eye involvement can reduce the risk of eye involvement in Behcet's patients. Therefore, reducing the onset time of disease symptoms and providing appropriate treatment can reduce Behcet's disease ocular complications.

24. A NOVEL TUMOR STAGING SYSTEM INCORPORATING CN STATUS FOR STRATIFYING EARLY STAGE ESOPHAGEAL SQUAMOUS CELL CARCINOMA PATIENTS AFTER TRIMODAL THERAPY

Abstract Backgrounds The aim of this study is to investigate the prognostic value of cN status for early stage esophageal squamous cell carcinoma (ESCC) patients after neoadjuvant chemoradiotherapy (nCRT) and construct a new staging model for individual survival prediction. Methods Patients from West China Hospital and Nanjing Jinling Hospital were included for training and validation, respectively. Patients with ESCC who underwent nCRT and esophagectomy were included in this study. Prognostic factors were investigated in patients with ypT0–3N0M0 using survival analysis and Cox regression model. Results For patients with ypI stage, cN status was a significant prognostic factor. The patients with cN1–3 yielded significantly worse OS (p = 0.014) and DFS (p = 0.008) compared with patients with cN0. However, for patients with ypII stage, cN status was not a significant prognostic factor (OS: p = 0.100 and DFS: p = 0.110). Cox regression analysis showed that only cN status was an independent prognostic factor for both OS (p = 0.030) and DFS (p = 0.019) in patients with stage ypI. A new staging model was established based on ypTNM stage and cN status by precisely stratifying ypI ESCC patients. Conclusion We constructed a new staging system based on ypTNM stage and cN status to precisely stratify the patients with ypI stage. Our new ypTNM-cN staging system provides new insights for classifying stage ypI ESCC and shows reliable classification efficacy for all ESCC patients after nCRT and surgery.

Cost-effectiveness of combination therapy of polatuzumab vedotin plus rituximab, cyclophosphamide, doxorubicin, and prednisolone for previously untreated diffuse large B-cell lymphoma in Japan

AimThe POLARIX trial showed that Pola + R-CHP (polatuzumab vedotin plus rituximab, cyclophosphamide, doxorubicin, and prednisolone) prolongs progression-free survival (PFS) in patients with previously untreated diffuse large B-cell lymphoma (DLBCL) compared with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone), the conventional standard of care, with a similar safety profile. However, Pola + R-CHP has not been evaluated from the viewpoint of health economics in Japan. This study evaluates the cost-effectiveness of Pola + R-CHP for previously untreated DLBCL from a Japanese public healthcare payer’s perspective.MethodsA partitioned survival analysis model was constructed to estimate lifetime costs and effectiveness of Pola + R-CHP and R-CHOP in previously untreated DLBCL who had an International Prognostic Index score (IPI) score of ≥2. A parametric survival model was applied to data analyzed in the POLARIX trial to estimate the lifetime overall survival (OS) and PFS for each treatment. The parameters required for the model were based on the results of a literature search and expert opinion.ResultsThe incremental cost-effectiveness ratio (ICER) of Pola + R-CHP vs. R-CHOP was JPY2,710,238 per quality-adjusted life year (QALY), less than the ICER of JPY7.5 million per QALY that is considered to be cost-effective based on the threshold of the Japanese cost-effectiveness evaluation system. One-way sensitivity analysis showed that the parameters influencing the results of the analysis were median PFS and the total cost per regimen of salvage chemotherapy, patient weight, and patient age. Probabilistic sensitivity analysis showed that the probability of Pola + R-CHP having superior cost-effectiveness was 99.2% when the reference value was JPY7.5 million. The results of scenario analysis suggested that prolongation of PFS was an important factor in the evaluation of cost-effectiveness in previously untreated DLBCL with or without prolongation of OS.ConclusionsThis study suggests that Pola + R-CHP is a cost-effective treatment for previously untreated DLBCL in Japan under the public health insurance system.