Sequential Shrinkage Estimate for COX Regression Models with Uncertain Number of Effective Variables

Abstract Objective To assess the relationships between female hormonal exposures and risk of rheumatoid arthritis (RA), in a prospective cohort of French women. Methods E3N is an on-going French prospective cohort that included 98 995 women aged 40–65 years in 1990. Every 2–3 years, women completed mailed questionnaires on their lifestyles, reproductive factors, and health conditions. Cox proportional-hazards regression models were used to determine factors associated with risk of incident RA, with age as the time scale, adjusted for known risk factors of RA, and considering endogenous and exogenous hormonal factors. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated. Effect modification by smoking history was investigated. Results A total of 698 incident cases of RA were ascertained among 78 452 women. In multivariable-adjusted Cox regression models, risk of RA was increased with early age at first pregnancy (<22 vs ≥27 years; HR = 1.34; 95%CI 1.0–1.7) and menopause (≤45 vs ≥53 years; HR = 1.40; 95%CI 1.0–1.9). For early menopause, the association was of similar magnitude in ever and never smokers, although the association was statistically significant only in ever smokers (HR = 1.54; 95%CI 1.0–2.3). We found a decreased risk in nulliparous women never exposed to smoking (HR = 0.44; 95%CI 0.2–0.8). Risk of RA was inversely associated with exposure to progestogen only in perimenopause (>24 vs 0 months; multi-adjusted HR = 0.77; 95%CI 0.6–0.9). Conclusions These results suggest an effect of both endogenous and exogenous hormonal exposures on RA risk and phenotype that deserves further investigation.

Download Full-text

Treatment Time or Convection Volume in HDF: What Drives the Reduced Mortality Risk?

Blood Purification ◽

10.1159/000430903 ◽

2015 ◽

Vol 40 (1) ◽

pp. 53-58 ◽

Cited By ~ 1

Author(s):

Camiel L.M. de Roij van Zuijdewijn ◽

Menso J. Nubé ◽

Piet M. ter Wee ◽

Peter J. Blankestijn ◽

Renée Lévesque ◽

...

Keyword(s):

Confidence Intervals ◽

Mortality Risk ◽

Survival Benefit ◽

Regression Models ◽

Cox Regression ◽

Treatment Time ◽

High Volume ◽

P Values ◽

Hazard Ratios

Background/Aims: Treatment time is associated with survival in hemodialysis (HD) patients and with convection volume in hemodiafiltration (HDF) patients. High-volume HDF is associated with improved survival. Therefore, we investigated whether this survival benefit is explained by treatment time. Methods: Participants were subdivided into four groups: HD and tertiles of convection volume in HDF. Three Cox regression models were fitted to calculate hazard ratios (HRs) for mortality of HDF subgroups versus HD: (1) crude, (2) adjusted for confounders, (3) model 2 plus mean treatment time. As the only difference between the latter models is treatment time, any change in HRs is due to this variable. Results: 114/700 analyzed individuals were treated with high-volume HDF. HRs of high-volume HDF are 0.61, 0.62 and 0.64 in the three models, respectively (p values <0.05). Confidence intervals of models 2 and 3 overlap. Conclusion: The survival benefit of high-volume HDF over HD is independent of treatment time.

Download Full-text

Kidney Measures with Diabetes and Hypertension on Cardiovascular Disease: The Atherosclerosis Risk in Communities Study

American Journal of Nephrology ◽

10.1159/000433450 ◽

2015 ◽

Vol 41 (4-5) ◽

pp. 409-417 ◽

Cited By ~ 6

Author(s):

Nadine Alexander ◽

Kunihiro Matsushita ◽

Yingying Sang ◽

Shoshana Ballew ◽

Bakhtawar K. Mahmoodi ◽

...

Keyword(s):

Cardiovascular Risk ◽

Regression Models ◽

Estimated Glomerular Filtration Rate ◽

Cox Regression ◽

Cardiovascular Outcomes ◽

Creatinine Ratio ◽

Atherosclerosis Risk In Communities ◽

Atherosclerosis Risk ◽

Lower Egfr

Background: Whether the association of chronic kidney disease (CKD) with cardiovascular risk differs based on diabetes mellitus (DM) and hypertension (HTN) status remains unanswered. Methods: We investigated 11,050 participants from the Atherosclerosis Risk in Communities Study (fourth examination (1996-1998)) with follow-up for cardiovascular outcomes (coronary disease, heart failure and stroke) through 2009. Using the Cox regression models, we quantified cardiovascular risk associated with estimated glomerular filtration rate (eGFR) and urinary albumin-to-creatinine ratio (ACR) in individuals with and without DM and/or HTN and assessed their interactions. Results: Individuals with DM and HTN generally had higher cardiovascular risk relative to those without at all the levels of eGFR and ACR. Cardiovascular risk increased with lower eGFR and higher ACR regardless of DM and HTN status (e.g. adjusted hazards ratio (HR) for eGFR 30-44 vs. 90-104 ml/min/1.73 m2, 2.32 (95% CI, 1.66-3.26) in non-diabetics vs. 1.83 (1.25-2.67) in diabetics and 2.45 (2.20-5.01) in non-hypertensives vs. 1.51 (1.27-1.81) in hypertensives and corresponding adjusted HR for ACR 30-299 vs. <10 mg/g, 1.70 (1.45-2.00) vs. 1.34 (1.10-1.64) and 1.42 (1.10-1.85) vs. 1.57 (1.36-1.81), respectively). Only the ACR-DM interaction reached significance, with a shallower relative risk gradient among diabetics than among non-diabetics (p = 0.02). Analysis of individual cardiovascular outcomes showed similar results. Conclusion: Although individuals with DM and HTN generally had higher cardiovascular risk relative to those without these complications, both low eGFR and high ACR were associated with cardiovascular diseases regardless of the presence or absence of DM and HTN. These findings reinforce the importance of CKD in cardiovascular outcomes.

Download Full-text

Risk of major chronic diseases in transport, rescue and security industries: a longitudinal register-based study

Occupational and Environmental Medicine ◽

10.1136/oemed-2021-107764 ◽

2021 ◽

pp. oemed-2021-107764

Author(s):

Kimmo Herttua ◽

Linda Juel Ahrenfeldt ◽

Tapio Paljarvi

Keyword(s):

Oral Cancer ◽

Chronic Diseases ◽

Respiratory Disease ◽

Regression Models ◽

Cox Regression ◽

Chronic Respiratory Disease ◽

Police Force ◽

Rescue Service ◽

Active Population ◽

Increased Risk

ObjectiveTo investigate the risk of hospitalisation for major chronic diseases across representative transport, rescue and security industries.MethodsWe performed a register-based study of 624 571 workers from six industries in Denmark between 2000 and 2005, followed up hospitalisation for chronic diseases up to 17 years, and compared with a 20% random sample of the economically active population.ResultsHR from the Cox regression models showed that seafarers had higher risk of lung cancer (men: 1.54, 95% CI 1.31 to 1.81; women: 1.63, 95% CI 1.13 to 2.36), and male seafarers had higher risk of diabetes (1.32, 95% CI 1.21 to 1.43) and oral cancer (1.51, 95% CI 1.21 to 1.88). Men and women in land transport had increased risk of diabetes (men: 1.68, 95% CI 1.63 to 1.73; women 1.55, 95% CI 1.40 to 1.71) and chronic respiratory disease (men: 1.21, 95% CI 1.16 to 1.25; women 1.42, 95% CI 1.32 to 1.53). Among women, a higher risk of gastrointestinal cancer was observed in aviation (1.53, 95% CI 1.23 to 1.89) and police force (1.29, 95% CI 1.01 to 1.65), oral cancer in defence forces (1.83, 95% CI 1.20 to 2.79), and chronic respiratory disease in rescue service (1.47, 95% CI 1.21 to 1.77), while men in defence forces, police force and rescue service had mainly lower risk of these chronic diseases.ConclusionsWe observed considerable health disparities from chronic diseases across transport, rescue and security industries, with workers in seafaring and land transport generally bearing the greatest relative burden.

Download Full-text

Impact of Assisted Peritoneal Dialysis Modality on Outcomes: A Cohort Study of the French Language Peritoneal Dialysis Registry

American Journal of Nephrology ◽

10.1159/000494664 ◽

2018 ◽

Vol 48 (6) ◽

pp. 425-433 ◽

Cited By ~ 7

Author(s):

Solène Guilloteau ◽

Thierry Lobbedez ◽

Sonia Guillouët ◽

Christian Verger ◽

Maxence Ficheux ◽

...

Keyword(s):

Peritoneal Dialysis ◽

Lower Risk ◽

Regression Models ◽

Cox Regression ◽

French Language ◽

Dialysis Modality ◽

Technique Survival ◽

Hazard Ratios ◽

Effect Of Pd ◽

Assisted Peritoneal Dialysis

Background: Patients on peritoneal dialysis (PD) can be assisted by a nurse or a family member and treated either by automated PD (APD) or continuous ambulatory PD (CAPD). The aim of this study was to evaluate the effect of PD modality and type of assistance on the risk of transfer to haemodialysis (HD) and on the peritonitis risk in assisted PD patients. Method: This was a retrospective study based on data from the French Language PD Registry. All adults starting assisted PD in France between 2006 and 2015 were included. Events of interest were transfer to HD, peritonitis and death. Cox regression models were used for statistical analysis. Results: Among the 12,144 incident patients who started PD in France during the study period, 6,167 were assisted. There were 5,060 nurse-assisted and 1,095 family-assisted PD patients. Overall, 5,171 were treated by CAPD and 996 by APD. In multivariate analysis, CAPD, compared to APD, was not associated with the risk of transfer to HD (cause specific hazard ratios [cs-HR] 0.96 [95% CI 0.84–1.09]). Patients on nurse-assisted PD had a lower risk of transfer to HD than family assisted PD patients (cs-HR 0.85 [95% CI 0.75–0.97]). Neither PD modality nor type of assistance were associated with peritonitis risk. Conclusions: In assisted PD, technique survival was not associated with PD modality. Nurse-assisted patients had a lower risk of transfer to HD than family assisted patients. Peritonitis risk was not influenced either by PD modality, or by type of assistance. Both APD and CAPD should be offered to assisted-PD patients.

Download Full-text

Analysis models for variables associated with breastfeeding duration

Revista Paulista de Pediatria ◽

10.1590/s0103-05822013000300006 ◽

2013 ◽

Vol 31 (3) ◽

pp. 306-314 ◽

Cited By ~ 3

Author(s):

Edson Theodoro dos S. Neto ◽

Eliana Zandonade ◽

Adauto Oliveira Emmerich

Keyword(s):

Logistic Regression ◽

Protective Factors ◽

Regression Model ◽

Odds Ratio ◽

Regression Models ◽

Adjusted Odds Ratio ◽

Cox Regression ◽

Bottle Feeding ◽

Breastfeeding Duration ◽

Factors Associated

OBJECTIVE To analyze the factors associated with breastfeeding duration by two statistical models. METHODS A population-based cohort study was conducted with 86 mothers and newborns from two areas primary covered by the National Health System, with high rates of infant mortality in Vitória, Espírito Santo, Brazil. During 30 months, 67 (78%) children and mothers were visited seven times at home by trained interviewers, who filled out survey forms. Data on food and sucking habits, socioeconomic and maternal characteristics were collected. Variables were analyzed by Cox regression models, considering duration of breastfeeding as the dependent variable, and logistic regression (dependent variables, was the presence of a breastfeeding child in different post-natal ages). RESULTS In the logistic regression model, the pacifier sucking (adjusted Odds Ratio: 3.4; 95%CI 1.2-9.55) and bottle feeding (adjusted Odds Ratio: 4.4; 95%CI 1.6-12.1) increased the chance of weaning a child before one year of age. Variables associated to breastfeeding duration in the Cox regression model were: pacifier sucking (adjusted Hazard Ratio 2.0; 95%CI 1.2-3.3) and bottle feeding (adjusted Hazard Ratio 2.0; 95%CI 1.2-3.5). However, protective factors (maternal age and family income) differed between both models. CONCLUSIONS Risk and protective factors associated with cessation of breastfeeding may be analyzed by different models of statistical regression. Cox Regression Models are adequate to analyze such factors in longitudinal studies.

Download Full-text

Dental restoration longevity among geriatric and adult special needs patients

10.1101/202069 ◽

2017 ◽

Cited By ~ 1

Author(s):

D. J. Caplan ◽

Y. Li ◽

W. Wang ◽

S. Kang ◽

L. Marchini ◽

...

Keyword(s):

Special Needs ◽

Regression Models ◽

Cox Regression ◽

Dental Treatment ◽

Treatment Options ◽

Dental Restoration ◽

University Of Iowa ◽

Restorative Treatment ◽

Dental Restorations ◽

Iowa College

AbstractThis study aimed to describe the survival trajectory of dental restorations placed in an outpatient population of geriatric and adult special needs patients over a 15-year span, with particular interest in longevity of subsequent restorations in teeth that received multiple restorations over time. Dental restorations of different types and sizes in patients age ≥65 years treated between 2000-14 at the University of Iowa, College of Dentistry were followed until they incurred an event (i.e., restoration replacement, extraction of the tooth, or endodontic treatment of the tooth). Survival analysis and extended Cox regression models were used to generate hazards ratios for selected predictor variables. A total of 9184 restorations were followed in 1551 unique patients. During the follow-up period, 28.7% of these restorations incurred an event; and overall the restorations had a median lifespan of 6.25 years. In multivariable regression models, after controlling for gender and age, composite restorations and greater number of restoration surfaces were associated with higher risks of failure; and the initial restoration recorded in the database for each subject tended to have lower risk of failure than restorations placed later that included any of those same surfaces. This information potentially could be helpful to elderly patients considering various restorative treatment options during the dental treatment planning and informed consent process.

Download Full-text

Identiﬁcation and Validation of Two Prognostic Autophagy-related Models in Clear Cell Renal Cell Carcinoma

10.21203/rs.3.rs-78750/v1 ◽

2020 ◽

Author(s):

Xiong-Lin Sun ◽

Jun-Ming Zhu ◽

Fei Lin ◽

Zhi-Bin Ke ◽

Xiao-Dong Li ◽

...

Keyword(s):

Renal Cell Carcinoma ◽

Regression Models ◽

Risk Model ◽

Cox Regression ◽

Clear Cell ◽

Risk Group ◽

Gene Model ◽

Cell Renal Cell Carcinoma ◽

Training Cohort ◽

Model Training

Abstract Background: Clear cell renal cell carcinoma (ccRCC) is one of the most frequent malignancies. Increasing evidence has highlighted the critical roles of autophagy-related genes and autophagy-related long non-coding RNA (lncRNA) in ccRCC development and progression. Therefore, it is necessary to identify novel biomarkers associated with autophagy in ccRCC.Methods: A total of 507 ccRCC patients were included in our study and then randomly divided into a training cohort (n=255) and testing cohort (n=252). Univariate Cox regression models, Lasso regression analyses and multivariate Cox regression models were successively used for constructing gene model and lncRNA model. Receiver-operating characteristic (ROC) curve analysis, Kaplan-Meier (K-M) analysis and more functional analyses were applied for verifying the accuracy of the two models.Results: The autophagy-related genes (ARGs) prognostic model was constructed based on the six ARGs (EIF4EBP1, IFNG, BID, BIRC5, CX3CL1 and RAB24) and five autophagy-related lncRNAs (AC093278.2, AC010326.3, AC099850.3, AC016773.1 and AC009549.1) and then ccRCC patients were significantly stratified into high- and low-risk groups in terms of overall survival (OS). K-M survival analyses indicated that low-risk group had a lower mortality rate than high-risk group in the six-gene prognostic risk model (training cohort: P=4.138e-07; testing cohort: P=1.125e-03) and the same results were obtained in the case of the five-lncRNA prognostic risk model (training cohort: P=4.564e-09; testing cohort: P=2.485e-03). The results of time-dependent ROC curves revealed six-gene prognostic risk model had a higher area under curve (AUC) of 0.765 than the five-lncRNA prognostic risk model at an AUC of 0.759. Therefore, the gene model is an indicator as good as the model constructed by lncRNAs. In addition, further functional analysis indicated these genes were functionally involved in regulation of endopeptidase activity, regulation of peptidase activity, autophagy, human cytomegalovirus infection, shigellosis, autophagy-animal and HIF-1 signaling pathway.Conclusions: A total of six OS-related ARGs and five autophagy-related lncRNA were identified in our current study. The two autophagy-related prediction models including genes and lncRNAs are reliable prognostic and predictive biomarkers for ccRCC.

Download Full-text

Effect of Berkson measurement error on parameter estimates in Cox regression models

Lifetime Data Analysis ◽

10.1007/s10985-007-9036-2 ◽

2007 ◽

Vol 13 (2) ◽

pp. 261-272 ◽

Cited By ~ 16

Author(s):

Helmut Küchenhoff ◽

Ralf Bender ◽

Ingo Langner

Keyword(s):

Measurement Error ◽

Regression Models ◽

Cox Regression ◽

Parameter Estimates

Download Full-text

Abstract P349: Association Of Fibroblast Growth Factor 23 With Incident Cardiovascular Disease And All-Cause Mortality In The Framingham Heart Study

Circulation ◽

10.1161/circ.129.suppl_1.p349 ◽

2014 ◽

Vol 129 (suppl_1) ◽

Author(s):

Robin Haring ◽

Ramachandran S Vasan ◽

Henri Wallaschofski ◽

Lisa Sullivan ◽

Danielle Enserro

Keyword(s):

Cardiovascular Disease ◽

Growth Factor ◽

Fibroblast Growth Factor ◽

Regression Models ◽

Cox Regression ◽

Fibroblast Growth Factor 23 ◽

Positive Association ◽

Fibroblast Growth ◽

All Cause Mortality ◽

Incident Cardiovascular Disease

Objective: To investigate the association of fibroblast growth factor 23 (FGF23) with incident cardiovascular disease (CVD) and mortality risk in the general population. Methods: We evaluated 3,236 Framingham Offspring and Omni Study participants to examine the associations of serum FGF23 (measured by immunoassay) with 10-year incident CVD (N = 2,823) and all-cause mortality (N = 3,223) using multivariable Cox regression models. Results: During a median follow-up time of 10.8 years (Q1, 10.0; Q3, 11.4), 347 participants developed new-onset CVD and 412 died. Age- and sex-adjusted Cox regression models revealed a positive association of FGF23 with incident CVD (hazard ratio (HR) per unit increase in logFGF23: 1.43, 95% confidence interval (CI) 1.11-1.84) and all-cause mortality (HR 2.26, 95% CI, 1.86-2.75). After multivariable adjustment, the association of FGF23 with incident CVD was rendered non-significant (HR 1.12, 95% CI 0.86-1.46), whereas the positive association of FGF23 with all-cause mortality was maintained (HR: 1.87, 95% CI: 1.52 - 2.29). Analyses modeling FGF23 quartiles yielded similar findings (multivariable-adjusted HR Q4 vs. Q1 for incident CVD: 1.17, 95% CI: 0.87 - 1.59; for death: 1.87, 95% CI: 1.38 - 2.53). Conclusion: In our large community-based sample, serum FGF23 shows an independent positive association with all-cause mortality, but not with incident CVD risk.

Download Full-text