Gaining more flexibility in Cox proportional hazards regression models with cubic spline functions
Abstract
The Cox proportional hazards model is the most popular model for the analysis of survival data. The use of cubic spline functions allows investigation of non-linear effects of continuous covariates and flexible assessment of time-by-covariate interactions. Two main advantages are provided—no particular functional form has to be specified and standard computer software packages like SAS or BMDP can be used. A SAS macro which implements the method is presented.
References (5)
- S. Durrleman et al.
Flexible regression models with cubic splines
Stat. Med.
(1989) - K.R. Hess
Assessing time-by-covariate interactions in proportional hazards regression models using cubic spline functions
Stat. Med.
(1994)
Cited by (255)
Bidirectional associations and a causal mediation analysis between depressive symptoms and chronic digestive diseases: A longitudinal investigation
2023, Journal of Affective DisordersChronic digestive diseases (CDDs) and depression shared major pathogeneses. We aimed to prospectively examine the bidirectional incidence associations between depressive symptoms and CDDs and explore biologically and behaviorally relevant mediators in the bidirectional associations.
Multivariable-adjusted Cox proportional hazard models were used to examine baseline depressive symptoms in relation to incident CDDs among 10,974 adults and the relation of baseline CDDs with new-onset elevated depressive symptoms among 7489 participants in the China Health and Retirement Longitudinal Study of nationally representative middle-aged and older adults. Elevated depressive symptoms were defined as the 10-item Center for Epidemiologic Studies Depression scale (CES-D-10) score at or higher than 10 and CDDs (except for tumor and cancer) were determined by self-reported physician diagnoses. Causal mediation analysis was performed to assess the mediated effects of a priori selected blood biomarkers and lifestyle factors in the bidirectional associations.
Prevalence of elevated depressive symptoms and nonmalignant CDDs at baseline was 33.05 % and 17.8 % respectively. During a mean of 5.47 years of follow-up, elevated depressive symptoms significantly increased hazard of CDDs by 1.66 folds (95%CI = 1.49–1.84). Having CDDs at baseline was associated with a 27 % (95%CI = 16 %–39 %) increased hazard of developing elevated depressive symptoms. Shorter sleeping duration at night nominally significantly mediated 8.76 % of the association between depressive symptoms and incident CDDs while no significant mediators were identified in the converse association.
Limited mediator information and inadequately long follow-up may reduce chance of identifying significant mediators.
Depressive symptoms and CDDs were mutual independent risk factors. Early screening and management of depressive symptoms and sleep disturbance are suggested in the prevention of CDDs and related comorbidities.
Combination Moderate-Intensity Statin and Ezetimibe Therapy for Elderly Patients With Atherosclerosis
2023, Journal of the American College of CardiologyThe routine use of high-intensity statins should be considered carefully in elderly patients because of their higher risk of intolerance or adverse events.
We evaluated the impact of moderate-intensity statin with ezetimibe combination therapy compared with high-intensity statin monotherapy in elderly patients with atherosclerotic cardiovascular disease (ASCVD).
In this post hoc analysis of the RACING (RAndomized Comparison of Efficacy and Safety of Lipid-lowerING With Statin Monotherapy Versus Statin/Ezetimibe Combination for High-risk Cardiovascular Diseases) trial, patients were stratified by age (≥75 years and <75 years). The primary endpoint was a 3-year composite of cardiovascular death, major cardiovascular events, or nonfatal stroke.
Among the 3,780 enrolled patients, 574 (15.2%) were aged ≥75 years. The rates of the primary endpoint were not different between the moderate-intensity statin with ezetimibe combination therapy group and the high-intensity statin monotherapy group among patients aged ≥75 years (10.6% vs 12.3%; HR: 0.87; 95% CI: 0.54-1.42; P = 0.581) and those <75 years (8.8% vs 9.4%; HR: 0.94; 95% CI: 0.74-1.18; P = 0.570) (P for interaction = 0.797). Moderate-intensity statin with ezetimibe combination therapy was associated with lower rates of intolerance-related drug discontinuation or dose reduction among patients aged ≥75 years (2.3% vs 7.2%; P = 0.010) and those <75 years (5.2% vs 8.4%; P < 0.001) (P for interaction = 0.159).
Moderate-intensity statin with ezetimibe combination therapy showed similar cardiovascular benefits to those of high-intensity statin monotherapy with lower intolerance-related drug discontinuation or dose reduction in elderly patients with ASCVD having a higher risk of intolerance, nonadherence, and discontinuation with high-intensity statin therapy. (RAndomized Comparison of Efficacy and Safety of Lipid-lowerING With Statin Monotherapy Versus Statin/Ezetimibe Combination for High-risk Cardiovascular Diseases [RACING Trial]; NCT03044665)
Elevated alanine transaminase is nonlinearly associated with in-hospital death in ICU-admitted diabetic ketoacidosis patients
2023, Diabetes Research and Clinical PracticeTo investigate the association between alanine transaminase (ALT) and in-hospital death in patients admitted to the intensive care unit for diabetic ketoacidosis (DKA).
A cohort of 2,684 patients was constructed from the eICU Collaborative Research Database. Baseline demographic and clinical characteristics were summarized. Cox regressions with restricted cubic spline functions were modelled to explore the association between alanine transaminase and in-hospital death. Subgroup analyses were conducted between sexes, age groups, and people with/without obesity.
After adjusting multiple confounders, a nonlinear, S-shaped association between ALT and in-hospital death was found. Compared to patients at median ALT, patients at the 90th percentile of ALT have a 1.88 (95 % confidence interval [CI]: 1.34–2.62) times higher hazard of in-hospital death in the unstratified cohort. Similar results were found in males (hazard ratio [HR] = 1.69, 95 % CI: 1.24–2.30); patients aged under 65 years (HR = 1.65, 95 % CI: 1.09–2.49); patients aged 65 years or above (HR = 3.45, 95 % CI: 1.67–7.14); non-obese patients (HR = 1.52, 95 % CI: 1.00–2.32); and obese patients (HR = 2.76, 95 % CI: 1.38–5.54).
Elevated ALT is robustly associated with in-hospital death in ICU-admitted DKA patients across several subgroups. Close monitoring of ALT in these patients is recommended.
Platelet Reactivity and Clinical Outcomes After Drug-Eluting Stent Implantation: Results From the PTRG-DES Consortium
2022, JACC: Cardiovascular InterventionsThe long-term prognostic implication of platelet reactivity after percutaneous coronary intervention (PCI) is not clearly known.
The impacts of platelet reactivity from the PTRG-DES consortium were assessed.
The primary endpoint was the major adverse cardiac and cerebrovascular events (MACCE) including all-cause death, myocardial infarction, stent thrombosis, or stroke. Key secondary endpoints were all-cause mortality, major bleeding, and net adverse clinical events (NACE), including MACCE and bleeding.
Between 2003 and 2018, a total of 11,714 patients were enrolled and grouped into tertiles according to P2Y12 reaction units (PRUs): high PRUs (≥253), intermediate PRUs (188-252), and low PRUs (<188). The Kaplan-Meier (KM) estimates of the primary outcome were significantly different across the groups; the high-PRU group showed the highest MACCE rate at 5 years (12.9%, 11.1%, and 7.0% in high-, intermediate-, and low-PRU groups, respectively; P < 0.001), as well as at 1 year (P < 0.001). The high-PRU group had the greatest KM estimates of all-cause death (8.2%, 5.9%, and 3.7%, respectively; P < 0.001) at 5 years without significant differences of major bleeding, and resultant of a higher KM estimates of NACE (15.7%, 13.6%, and 9.7%, respectively; P < 0.001). A PRU ≥252, the best cutoff value, was strongly related to MACCE (HR: 1.39; 95% CI: 1.11-1.74; P = 0.003) and all-cause death at 5 years after PCI (HR: 1.42; 95% CI: 1.04-1.94; P = 0.026). The optimal cutoff value of aspirin reaction units predicting the MACCE occurrence was ≥414 and was significantly associated with 5-year MACCE occurrence or all-cause death (P < 0.001).
In this large-scale cohort, high PRU was significantly associated with occurrence of MACCE, all-death death, and NACE at 5 years, as well as 1 year after PCI. (PTRG-DES Consortium [PTRG]; NCT04734028)
Factor VIII/protein C ratio independently predicts liver-related events but does not indicate a hypercoagulable state in ACLD
2022, Journal of HepatologyIt has been suggested that the ratio of procoagulant factor VIII to anticoagulant protein C (FVIII/PC) reflects the hemostatic equilibrium. Moreover, FVIII/PC predicted decompensation/death in a small study not accounting for portal hypertension severity. We investigated (i) the prognostic value of FVIII/PC (outcome-cohort) and (ii) whether FVIII/PC reflects the hypercoagulable state (assessed by thrombomodulin-modified thrombin generation assay [TM-TGA]) or the risk of bleeding/thrombotic events in patients undergoing hepatic venous pressure gradient (HVPG) measurement during follow-up.
(i) The outcome-cohort comprised 576 patients with evidence of advanced chronic liver disease (liver stiffness measurement ≥10 kPa and/or HVPG ≥6 mmHg). (ii) TM-TGA-cohort patients (n = 142) were recruited from the prospective VIenna CIrrhosis Study (VICIS: NCT03267615).
(i) FVIII/PC significantly increased across clinical stages (p <0.001) as well as HVPG (p <0.001) and MELD score (p <0.001) strata and remained independently associated with decompensation/liver-related death (adjusted hazard ratio 1.06; 95% CI 1.01–1.11; p = 0.013), even after multivariable adjustment. It was also associated with acute-on-chronic liver failure (ACLF) development (adjusted hazard ratio 1.10; 95% CI 1.02-1.19; p = 0.015) in patients with decompensated cirrhosis. (ii) FVIII/PC showed a weak positive correlation with endogenous thrombin potential (Spearman’s ρ = 0.255; p = 0.002), but this association disappeared after adjusting for the severity of liver disease. FVIII/PC was not associated with the development of bleeding (p = 0.272) or thrombotic events (p = 0.269). However, FVIII/PC correlated with biomarkers of different pathophysiological mechanisms that promote liver disease progression.
FVIII/PC provides prognostic information regarding hepatic decompensation/death and ACLF, independently of established prognostic indicators. However, this is not evidence that hypercoagulability drives disease progression, as the correlation between FVIII/PC and thrombin generation is confounded by liver disease severity and FVIII/PC was not associated with thrombosis. Therefore, FVIII/PC does not reflect coagulation and results from previous studies on FVIII/PC require re-interpretation.
NCT03267615 (in part).
A balanced coagulation system is essential for preventing bleeding episodes and blood clot formation (thrombosis). Blood of patients with advanced liver disease may have increased coagulation potential, possibly promoting the worsening of liver disease via thrombosis in the blood vessels of the liver. The ratio between the results of 2 blood tests (procoagulant factor VIII to anticoagulant protein C) has been suggested to reflect these increases in coagulation potential. Our study demonstrates, on the one hand, that this ratio is a versatile predictor of the development of complications of cirrhosis, yet on the other hand, that it is unrelated to coagulation.
Association between exposure to parabens and total mortality in US adults
2022, Environmental ResearchParabens are a group of endocrine disruptors that have been associated with health effects such as hypertension, diabetes, oxidative stress and obesity, which are associated with increased mortality risk over time. Women are exposed to higher paraben levels than men through use of consumer products. The current prospective study examines paraben exposure in association with mortality risk for women and men.
We analyzed 2005–2008 National Health and Nutrition Examination Survey (NHANES) data on urinary paraben analyte concentrations and covariates in adults aged 20 years and over, prospectively linked to National Center for Health Statistics (NCHS) mortality through 2015 (N = 2939). Proportional hazard regression models examined mortality risk in association with exposures, controlling for covariates for women and men. Covariates included age, race/ethnicity, smoking, education, income, body mass index, physical activity, self-reported health status and baseline health conditions.
Women were exposed to significantly higher concentrations of all studied parabens than men. Exposures were highest for methyl paraben. Women had significantly higher mortality risk in association with higher natural log exposure to ethyl (HR = 2.048, 95% CI 1.164–3.601), methyl (HR = 1.312, 95% CI 1.013–1.700), butyl (HR = 2.719, 95% CI 1.591–4.647) and total parabens (HR = 1.292, 95% CI 1.006–1.659). Exposure concentrations were associated with higher mortality risk for men only for ethyl paraben (HR = 2.532, 95% CI 1.217–5.268).
Women were found to be at greater mortality risk in association with exposure to ethyl, methyl, butyl and total parabens. These findings require confirmatory research but add to the evidence base that exposure to parabens, probably through consumer products, may have adverse effects on human health, especially for women.