On the unification of long-term survival models

In this paper we extend the long-term survival model proposed by Chen et al. [Chen, M.-H., Ibrahim, J.G., Sinha, D., 1999. A new Bayesian model for survival data with a surviving fraction. journal of the American Statistical Association 94, 909-919] via the generating function of a real sequence introduced by Feller [Feller, W., 1968. An Introduction to Probability Theory and its Applications, third ed., vol. 1, Wiley, New York]. A direct consequence of this new formulation is the unification of the long-term survival models proposed by Berkson and Gage [Berkson, J., Gage, R.P., 1952. Survival cure for cancer patients following treatment. journal of the American Statistical Association 47, 501-515] and Chen et al. (see citation above). Also, we show that the long-term survival function formulated in this paper satisfies the proportional hazards property if, and only if, the number of competing causes related to the occurrence of an event of interest follows a Poisson distribution. Furthermore, a more flexible model than the one proposed by Yin and Ibrahim [Yin, G., Ibrahim, J.G., 2005. Cure rate models: A unified approach. The Canadian journal of Statistics 33, 559-570] is introduced and, motivated by Feller`s results, a very useful competing index is defined. (c) 2008 Elsevier B.V. All rights reserved.

Recidiva de câncer labial em pacientes atendidos no hospital DR. Luiz Antônio (Natal-RN) entre 1997-2004

Introduction: Cancer of the lip is very common in tropical countries, being noticeable the squamous cell carcinoma as the main histological type. Objective: Evaluate the socialdemographic profile, habits, occupation, clinical characteristics of the cancer lesions and the aftermath of treatment of the patients treated on the Luiz Antônio Hospital (Natal-RN). Design: Retrospective cohort. Methods: We analyzed 181 medical records of patients from the Luiz Antônio Cancer Hospital (Natal-RN) in the period between 1997 and 2004. The statistic evaluation of time between the diagnosis and the relapse or the cure of the patient were done through the Kaplan-Meier method and the comparison of survivor functions were done through the Log-rank test. Later, was estimated the proportional risk model of Cox. Results: The study population were composed by 69,1% males, 95,2% unlettered, the mean age of 66,5 years, 89,0% of smokers and 64,1% had an occupation involving sun exposure. In regard to the clinical characteristics, most lesions were in the lower lip (77,9%), the size of the tumor was smaller than 2 cm (51,8%), 92,6% had localized lesions. Were verified 16,3% of local relapse and 13% of regional. Almost the totally of the cases corresponded to squamous cell carcinoma (97,2%). We observed smaller accumulated probability of not occurrence of local relapse when the base and borders were free of lesions (p=0,041), as well as a smaller probability of regional relapse when the sort of treatment was surgery, associated with other therapeutics modalities (p=0,001). The patients with advanced pathologic stage (p=0,016), treated with surgery associated with other therapeutics modalities (p=0,001) and diameter above 4cm (p=0,019) presented a bigger possibility of any kind of relapse. The multivariable analysis pointed the complex treatments (surgery plus other therapeutics modalities) as a predictor variable for occurrence of new local lesions (p=0,001) and total (p=0,046), besides the age above 70 years to the regional relapse (p=0,050). Conclusion: Cancer of the lip occur in the lower lip, in males, smokers and individuals exposed to Sun light. The relapse was frequent, even being localized and without great consequences to the patient s health. The probability of relapse is related to the size and borders of the lesion and to the histological exam, as well as to the patient s age and complexity of the treatment chosen

Geographic and Educational Factors and Risk of the First Peritonitis Episode in Brazilian Peritoneal Dialysis Study (BRAZPD) Patients

Background and objectives Peritonitis remains as the most frequent cause of peritoneal dialysis (PD) failure, impairing patient's outcome. No large multicenter study has addressed socioeconomic, educational, and geographic issues as peritonitis risk factors in countries with a large geographic area and diverse socioeconomic conditions, such as Brazil.Design, setting, participants, & measurements Incident PD patients recruited from 114 dialysis centers and reporting to BRAZPD, a multicenter observational study, from December 2004 through October 2007 were included. Clinical, dialysis-related, demographic, and socioeconomic variables were analyzed. Patients were followed up until their first peritonitis. Cox proportional model was used to determine independent factors associated with peritonitis.Results In a cumulative follow-up of 2032 patients during 22.026 patient-months, 474 (23.3%) presented a first peritonitis episode. In contrast to earlier findings, PD modality, previous hemodialysis, diabetes, gender, age, and family income were not risk predictors. Factors independently associated with increased hazard risk were lower educational level, non-white race, region where patients live, shorter distance from dialysis center, and lower number of patients per center.Conclusions Educational level and geographic factors as well as race and center size are associated with risk for the first peritonitis, independent of socioeconomic status, PD modality, and comorbidities. Clin J Am Soc Nephrol 6: 1944-1951, 2011. doi: 10.2215/CJN.11431210

Prognostic factors for survival in patients with colorectal liver metastases: experience of a single brazilian cancer center

CONTEXT: Liver metastases are a common event in the clinical outcome of patients with colorectal cancer and account for 2/3 of deaths from this disease. There is considerable controversy among the data in the literature regarding the results of surgical treatment and prognostic factors of survival, and no analysis have been done in a large cohort of patients in Brazil. OBJECTIVES: To characterize the results of surgical treatment of patients with colorectal liver metastases, and to establish prognostic factors of survival in a Brazilian population. METHOD: This was a retrospective study of patients undergoing liver resection for colorectal metastases in a tertiary cancer hospital from 1998 to 2009. We analyzed epidemiologic variables and the clinical characteristics of primary tumors, metastatic disease and its treatment, surgical procedures and follow-up, and survival results. Survival analyzes were done by the Kaplan-Meier method and the log-rank test was applied to determine the influence of variables on overall and disease-free survival. All variables associated with survival with P<0.20 in univariate analysis, were included in multivariate analysis using a Cox proportional hazard regression model. RESULTS: During the period analyzed, 209 procedures were performed on 170 patients. Postope-rative mortality in 90 days was 2.9% and 5-year overall survival was 64.9%. Its independent prognostic factors were the presence of extrahepatic disease at diagnosis of liver metastases, bilateral nodules and the occurrence of major complications after liver surgery. The estimated 5-year disease-free survival was 39.1% and its prognostic factors included R1 resection, extrahepatic disease, bilateral nodules, lymph node involvement in the primary tumor and primary tumors located in the rectum. CONCLUSION: Liver resection for colorectal metastases is safe and effective and the analysis of prognostic factors of survival in a large cohort of Brazilian patients showed similar results to those pointed in international series. The occurrence of major postoperative complications appears to be able to compromise overall survival and further investigation in needed in this topic.

Non-fatal cardiovascular outcome in patients with posttraumatic stress symptoms caused by myocardial infarction

Objectives Posttraumatic stress disorder (PTSD) prospectively increases the risk of incident cardiovascular disease (CVD) independent of other risk factors in otherwise healthy individuals. Between 10% and 20% of patients develop PTSD related to the traumatic experience of myocardial infarction (MI). We investigated the hypothesis that PTSD symptoms caused by MI predict adverse cardiovascular outcome. Methods We studied 297 patients (61 ± 10 years, 83% men) who self-rated PTSD symptoms attributable to a previous index MI. Non-fatal CVD-related hospital readmissions (i.e. recurrent MI, elective and non-elective intracoronary stenting, bypass surgery, pacemaker implantation, cardiac arrhythmia, cerebrovascular event) were assessed at follow-up. Cox proportional hazard models controlled for demographic factors, coronary heart disease severity, major CVD risk factors, cardiac medication, and mental health treatment. Results Forty-three patients (14.5%) experienced an adverse event during a mean follow-up of 2.8 years (range 1.3–3.8). A 10 point higher level in the PTSD symptom score (mean 8.8 ± 9.0, range 0–47) revealed a hazard ratio (HR) of 1.42 (95% CI 1.07–1.88) for a CVD-related hospital readmission in the fully adjusted model. A similarly increased risk (HR 1.45, 95% CI 1.07–1.97) emerged for patients with a major or unscheduled CVD-related readmission (i.e. when excluding patients with elective stenting). Conclusions Elevated levels of PTSD symptoms caused by MI may adversely impact non-fatal cardiovascular outcome in post-MI patients independent of other important prognostic factors. The possible importance of PTSD symptoms as a novel prognostic psychosocial risk factor in post-MI patients warrants further study.

Designed Extension of Survival Studies: Application to Clinical Trials with Unrecognized Heterogeneity

It is well known that unrecognized heterogeneity among patients, such as is conferred by genetic subtype, can undermine the power of randomized trial, designed under the assumption of homogeneity, to detect a truly beneficial treatment. We consider the conditional power approach to allow for recovery of power under unexplained heterogeneity. While Proschan and Hunsberger (1995) confined the application of conditional power design to normally distributed observations, we consider more general and difficult settings in which the data are in the framework of continuous time and are subject to censoring. In particular, we derive a procedure appropriate for the analysis of the weighted log rank test under the assumption of a proportional hazards frailty model. The proposed method is illustrated through application to a brain tumor trial.

Additive Hazards Models with Latent Treatment Effectiveness Lag Time

In many clinical trials to evaluate treatment efficacy, it is believed that there may exist latent treatment effectiveness lag times after which medical procedure or chemical compound would be in full effect. In this article, semiparametric regression models are proposed and studied to estimate the treatment effect accounting for such latent lag times. The new models take advantage of the invariance property of the additive hazards model in marginalizing over random effects, so parameters in the models are easy to be estimated and interpreted, while the flexibility without specifying baseline hazard function is kept. Monte Carlo simulation studies demonstrate the appropriateness of the proposed semiparametric estimation procedure. Data collected in the actual randomized clinical trial, which evaluates the effectiveness of biodegradable carmustine polymers for treatment of recurrent brain tumors, are analyzed.

Modeling multilevel sleep transitional data via Poisson log-linear multilevel models

This paper proposes Poisson log-linear multilevel models to investigate population variability in sleep state transition rates. We specifically propose a Bayesian Poisson regression model that is more flexible, scalable to larger studies, and easily fit than other attempts in the literature. We further use hierarchical random effects to account for pairings of individuals and repeated measures within those individuals, as comparing diseased to non-diseased subjects while minimizing bias is of epidemiologic importance. We estimate essentially non-parametric piecewise constant hazards and smooth them, and allow for time varying covariates and segment of the night comparisons. The Bayesian Poisson regression is justified through a re-derivation of a classical algebraic likelihood equivalence of Poisson regression with a log(time) offset and survival regression assuming piecewise constant hazards. This relationship allows us to synthesize two methods currently used to analyze sleep transition phenomena: stratified multi-state proportional hazards models and log-linear models with GEE for transition counts. An example data set from the Sleep Heart Health Study is analyzed.

Domestic radon exposure and risk of childhood cancer: a prospective census-based cohort study

Background: In contrast with established evidence linking high doses of ionizing radiation with childhood cancer, research on low-dose ionizing radiation and childhood cancer has produced inconsistent results. Objective: We investigated the association between domestic radon exposure and childhood cancers, particularly leukemia and central nervous system (CNS) tumors. Methods: We conducted a nationwide census-based cohort study including all children < 16 years of age living in Switzerland on 5 December 2000, the date of the 2000 census. Follow-up lasted until the date of diagnosis, death, emigration, a child’s 16th birthday, or 31 December 2008. Domestic radon levels were estimated for each individual home address using a model developed and validated based on approximately 45,000 measurements taken throughout Switzerland. Data were analyzed with Cox proportional hazard models adjusted for child age, child sex, birth order, parents’ socioeconomic status, environmental gamma radiation, and period effects. Results: In total, 997 childhood cancer cases were included in the study. Compared with children exposed to a radon concentration below the median (< 77.7 Bq/m3), adjusted hazard ratios for children with exposure ≥ the 90th percentile (≥ 139.9 Bq/m3) were 0.93 (95% CI: 0.74, 1.16) for all cancers, 0.95 (95% CI: 0.63, 1.43) for all leukemias, 0.90 (95% CI: 0.56, 1.43) for acute lymphoblastic leukemia, and 1.05 (95% CI: 0.68, 1.61) for CNS tumors. Conclusions: We did not find evidence that domestic radon exposure is associated with childhood cancer, despite relatively high radon levels in Switzerland.

The prognostic significance of sialyl-Tn antigen in women treated with adjuvant chemotherapy for breast cancer

Background and purpose. Sialyl-Tn(STn) represents an aberrantly glycosylated mucin epitope which is expressed in breast cancer and other adenocarcinomas and is an important target for the development of novel immunotherapeutic approaches. It is a marker of adverse prognosis in colon and ovarian cancer, but information about its prognostic impact in breast cancer is limited. The primary aim of the present study was to investigate the influence of STn expression on outcome of invasive breast cancer in 207 women who received anthracyline-containing adjuvant chemotherapy in a prospective clinical trial.^ Methods. Expression of STn was determined by an immunohistochemical procedure using the B72.3 monoclonal antibody. The extent of staining was determined by two observers using a 0 through 4 point scale, with 0 representing $<$5% of cells staining; 1: 5-25%; 2: 26-50%; 3: 51-75%; and 4: $>$75%. Intraobserver and interobserver agreement was.78-.92 (kappa). Kaplan-Meier and Cox proportional regression survival analyses were used to compare STn-negative and STn-positive patients.^ Results. Forty-eight (23%) of the 207 specimens demonstrated positive staining of STn. With a median follow-up of five years, STn-positivity was associated with a higher 5-year recurrence-free survival time than STn-negativity (67% vs. 80%, respectively; p = 0.03). STn expression was significantly associated with menopausal status (p = 0.04) but not other conventional prognostic markers. The risk of breast cancer recurrence and death was assessed by multivariate Cox regression analyses with adjustment for lymph node status, tumor size, menopausal status, hormone receptor status, nuclear grade, S-phase fraction and ploidy. In the final multivariate model for recurrence-free survival, the three factors that showed prognostic significance were: lymph node status (hazard ratio (HR) 3.04, 95% confidence interval (CI) 1.08-8.49), STn expression (HR 2.02, 95% CI 1.09-3.73), and tumor size (HR 1.96, 95% CI 1.05-3.64). STn was also associated with worse overall survival (HR 2.16, 95% CI 0.95-4.92) in multivariate analysis.^ Conclusion. STn antigen was shown to be a predictor of poor outcome in breast cancer. This tumor-associated antigen may be a valuable marker for identifying individuals at high risk of developing recurrent disease who may benefit from adjuvant therapy targeted at STn following definitive local therapy. Further study is needed to clarify the biologic and prognostic role of STn in breast cancer. ^

Evaluation of 9 biomarkers for predicting 10-year cardiovascular risk in patients undergoing coronary angiography: findings from the LUdwigshafen RIsk and Cardiovascular Health (LURIC) study.

BACKGROUND Conventional factors do not fully explain the distribution of cardiovascular outcomes. Biomarkers are known to participate in well-established pathways associated with cardiovascular disease, and may therefore provide further information over and above conventional risk factors. This study sought to determine whether individual and/or combined assessment of 9 biomarkers improved discrimination, calibration and reclassification of cardiovascular mortality. METHODS 3267 patients (2283 men), aged 18-95 years, at intermediate-to-high-risk of cardiovascular disease were followed in this prospective cohort study. Conventional risk factors and biomarkers were included based on forward and backward Cox proportional stepwise selection models. RESULTS During 10-years of follow-up, 546 fatal cardiovascular events occurred. Four biomarkers (interleukin-6, neutrophils, von Willebrand factor, and 25-hydroxyvitamin D) were retained during stepwise selection procedures for subsequent analyses. Simultaneous inclusion of these biomarkers significantly improved discrimination as measured by the C-index (0.78, P = 0.0001), and integrated discrimination improvement (0.0219, P<0.0001). Collectively, these biomarkers improved net reclassification for cardiovascular death by 10.6% (P<0.0001) when added to the conventional risk model. CONCLUSIONS In terms of adverse cardiovascular prognosis, a biomarker panel consisting of interleukin-6, neutrophils, von Willebrand factor, and 25-hydroxyvitamin D offered significant incremental value beyond that conveyed by simple conventional risk factors.

Variability in the Treatment of Elderly Patients with Stage IIIA (N2) Non–Small-Cell Lung Cancer

INTRODUCTION: We evaluated treatment patterns of elderly patients with stage IIIA (N2) non-small-cell lung cancer (NSCLC). METHODS: The use of surgery, chemotherapy, and radiation for patients with stage IIIA (T1-T3N2M0) NSCLC in the Surveillance, Epidemiology, and End Results-Medicare database from 2004 to 2007 was analyzed. Treatment variability was assessed using a multivariable logistic regression model that included treatment, patient, tumor, and census track variables. Overall survival was analyzed using the Kaplan-Meier approach and Cox proportional hazard models. RESULTS: The most common treatments for 2958 patients with stage IIIA (N2) NSCLC were radiation with chemotherapy (n = 1065, 36%), no treatment (n = 534, 18%), and radiation alone (n = 383, 13%). Surgery was performed in 709 patients (24%): 235 patients (8%) had surgery alone, 40 patients (1%) had surgery with radiation, 222 patients had surgery with chemotherapy (8%), and 212 patients (7%) had surgery, chemotherapy, and radiation. Younger age (p < 0.0001), lower T-status (p < 0.0001), female sex (p = 0.04), and living in a census track with a higher median income (p = 0.03) predicted surgery use. Older age (p < 0.0001) was the only factor that predicted that patients did not get any therapy. The 3-year overall survival was 21.8 ± 1.5% for all patients, 42.1 ± 3.8% for patients that had surgery, and 15.4 ± 1.5% for patients that did not have surgery. Increasing age, higher T-stage and Charlson Comorbidity Index, and not having surgery, radiation, or chemotherapy were all risk factors for worse survival (all p values < 0.001). CONCLUSIONS: Treatment of elderly patients with stage IIIA (N2) NSCLC is highly variable and varies not only with specific patient and tumor characteristics but also with regional income level.

Occupational exposure to magnetic fields and electric shocks and risk of ALS: The Swiss National Cohort

Amyotrophic lateral sclerosis (ALS) has been associated with exposures in so-called 'electrical occupations'. It is unclear if this possible link may be explained by exposure to extremely low-frequency magnetic fields (ELF-MF) or by electrical shocks. We evaluated ALS mortality in 2000-2008 and exposure to ELF-MF and electrical shocks in the Swiss National Cohort, using job exposure matrices for occupations at censuses 1990 and 2000. We compared 2.2 million workers with high or medium vs. low exposure to ELF-MF and electrical shocks using Cox proportional hazard models. Results showed that mortality from ALS was higher in people who had medium or high ELF-MF exposure in both censuses (HR 1.55 (95% CI 1.11-2.15)), but closer to unity for electrical shocks (HR 1.17 (95% CI 0.83-1.65)). When both exposures were included in the same model, the HR for ELF-MF changed little (HR 1.56), but the HR for electric shocks was attenuated to 0.97. In conclusion, there was an association between exposure to ELF-MF and mortality from ALS among workers with a higher likelihood of long-term exposure.

A Swiss paradox? Higher income inequality of municipalities is associated with lower mortality in Switzerland.

It has long been surmised that income inequality within a society negatively affects public health. However, more recent studies suggest there is no association, especially when analyzing small areas. This study aimed to evaluate the effect of income inequality on mortality in Switzerland using the Gini index on municipality level. The study population included all individuals >30 years at the 2000 Swiss census (N = 4,689,545) living in 2,740 municipalities with 35.5 million person-years of follow-up and 456,211 deaths over follow-up. Cox proportional hazard regression models were adjusted for age, gender, marital status, nationality, urbanization, and language region. Results were reported as hazard ratios (HR) with 95 % confidence intervals. The mean Gini index across all municipalities was 0.377 (standard deviation 0.062, range 0.202-0.785). Larger cities, high-income municipalities and tourist areas had higher Gini indices. Higher income inequality was consistently associated with lower mortality risk, except for death from external causes. Adjusting for sex, marital status, nationality, urbanization and language region only slightly attenuated effects. In fully adjusted models, hazards of all-cause mortality by increasing Gini index quintile were HR = 0.99 (0.98-1.00), HR = 0.98 (0.97-0.99), HR = 0.95 (0.94-0.96), HR = 0.91 (0.90-0.92) compared to the lowest quintile. The relationship of income inequality with mortality in Switzerland is contradictory to what has been found in other developed high-income countries. Our results challenge current beliefs about the effect of income inequality on mortality on small area level. Further investigation is required to expose the underlying relationship between income inequality and population health.

Calcium Disorders in the Emergency Department: Independent Risk Factors for Mortality.

BACKGROUND Calcium disorders are common in both intensive care units and in patients with chronic kidney disease and are associated with increased morbidity and mortality. It is unknown whether calcium abnormalities in unselected emergency department admissions have an impact on in-hospital mortality. METHODS This cross-sectional analysis included all admissions to the Emergency Department at the Inselspital Bern, Switzerland from 2010 to 2011. For hyper- and hypocalcaemic patients with a Mann-Whitney U-test, the differences between subgroups divided by age, length of hospital stay, creatinine, sodium, chloride, phosphate, potassium and magnesium were compared. Associations between calcium disorders and 28-day in-hospital mortality were assessed using the Cox proportional hazard regression model. RESULTS 8,270 patients with calcium measurements were included in our study. Overall 264 (3.2%) patients died. 150 patients (6.13%) with hypocalcaemia and 7 patients with hypercalcaemia (6.19%) died, in contrast to 104 normocalcaemic patients (1.82%). In univariate analysis, calcium serum levels were associated with sex, mortality and pre-existing diuretic therapy (all p<0.05). In multivariate Cox regression analysis, hypocalcaemia and hypercalcaemia were independent risk factors for mortality (HR 2.00 and HR 1.88, respectively; both p<0.01). CONCLUSION Both hypocalcaemia and hypercalcaemia are associated with increased 28-day in-hospital mortality in unselected emergency department admissions.