Working‐correlation‐structure identification in generalized estimating equations

Selecting an appropriate working correlation structure is pertinent to clustered data analysis using generalized estimating equations (GEE) because an inappropriate choice will lead to inefficient parameter estimation. We investigate the well‐known criterion of QIC for selecting a working correlation structure, and have found that performance of the QIC is deteriorated by a term that is theoretically independent of the correlation structures but has to be estimated with an error. This leads us to propose a correlation information criterion (CIC) that substantially improves the QIC performance. Extensive simulation studies indicate that the CIC has remarkable improvement in selecting the correct correlation structures. We also illustrate our findings using a data set from the Madras Longitudinal Schizophrenia Study. Copyright © 2008 John Wiley & Sons, Ltd.     

Assessing ethical problem solving by reasoning rather than decision making

Context The assessment of ethical problem solving in medicine has been controversial and challenging. The purposes of this study were: (i) to create a new instrument to measure doctors’ decisions on and reasoning approach towards resolving ethical problems; (ii) to evaluate the scores generated by the new instrument for their reliability and validity, and (iii) to compare doctors’ ethical reasoning abilities between countries and among medical students, residents and experts. Methods This study used 15 clinical vignettes and the think‐aloud method to identify the processes and components involved in ethical problem solving. Subjects included volunteer ethics experts, postgraduate Year 2 residents and pre‐clerkship medical students. The interview data were coded using the instruments of the decision score and Ethical Reasoning Inventory (ERI). The ERI assessed the quality of ethical reasoning for a particular case (Part I) and for an individual globally across all the vignettes (Part II). Results There were 17 Canadian and 32 Taiwanese subjects. Based on the Canadian standard, the decision scores between Taiwanese and Canadian subjects differed significantly, but made no discrimination among the three levels of expertise. Scores on the ERI Parts I and II, which reflect doctors’ reasoning quality, differed between countries and among different levels of expertise in Taiwan, providing evidence of construct validity. In addition, experts had a greater organised knowledge structure and considered more relevant variables in the process of arriving at ethical decisions than did residents or students. The reliability of ERI scores was 0.70–0.99 on Part I and 0.75–0.80 on Part II. Conclusions Expertise in solving ethical problems could not be differentiated by the decisions made, but could be differentiated according to the reasoning used to make those decisions. The difference between Taiwanese and Canadian experts suggests that cultural considerations come into play in the decisions that are made in the course of providing humane care to patients.     

Lung Cancer and COPD: a Common Combination

Background and objectiveTo analyse frequency, characteristics and patient survival with lung cancer (LC) and Common Obstructive Pulmonary Disease (COPD), comparing them with patients that do not have COPD.Material and methodsA retrospective study, of patients diagnosed by means of cytohistology. Survival was estimated by the Kaplan-Meier method. Statistical analysis was carried out using SPSS 15.0.ResultsA total of 996 patients were diagnosed, 39.8% with COPD. Mean age70±9.19 years. GOLD stages: I 18.2%, II 53.6%, III 24%, IV 4.2%. The histological types: squamous cell carcinoma 48.2%, adenocarcinoma 22%, and small cell carcinoma 22.5%. Survival was longer in the COPD group.ConclusionsLC and COPD are combined in 39.8%. Squamous cell type is more frequent and survival was longer in the COPD group.     

Monotherapy with enoxaparin for the prevention of recurrent venous thromboembolism.

This study aimed to determine whether a weight-adjusted dose of subcutaneous enoxaparin is as effective and safe as oral acenocoumarol for the secondary prophylaxis of pulmonary embolism. Three hundred and eighty consecutive noncancer outpatients hospitalized with an episode of symptomatic pulmonary embolism selected treatment with acenocoumarol or enoxaparin at a dose of 1 mg/kg once daily after being informed of the type of administration and expected frequency of laboratory monitoring for both medicinal products. Endpoints were symptomatic recurrent thromboembolic events evaluated by standard objective testing, and a composite endpoint of recurrent venous thromboembolism, major bleeding, and death from any cause. One hundred and ninety-nine patients (52%) chose acenocoumarol therapy and 181 chose enoxaparin monotherapy. Four patients in the enoxaparin group (2.2%) and six patients in the acenocoumarol group (3%) had an objective thromboembolic recurrence (hazard ratio, 1.35; 95% confidence interval, 0.38-4.79; P = 0.64). Nine patients in the enoxaparin group (5.0%) had a hemorrhagic complication compared with 11 in the acenocoumarol group (5.5%) (P = 0.81). The hospital length of stay was shorter with enoxaparin compared with acenocoumarol (11 versus 16 days, P = 0.0001). Enoxaparin is as effective and safe as acenocoumarol in the secondary prevention of recurrent thromboembolic disease and is associated with shorter hospitalization.     

Diagnostic value of D-dimer in outpatients with suspected deep venous thrombosis receiving oral anticoagulation.

D-dimer has proved a useful diagnostic tool for the exclusion of deep venous thrombosis (DVT). The objective of this paper was to evaluate the diagnostic performance of a diagnostic algorithm combining clinical probability and D-dimer in outpatients receiving oral anticoagulant treatment (OAT) similar to those regularly applied to nonanticoagulated individuals. We enrolled 70 outpatients on OAT who presented with clinically suspected DVT; a standard diagnostic algorithm including clinical evaluation using the modified Wells score and a quantitative immunoturbidimetric D-dimer assay (STA Liatest D-Di; Diagnostica Stago, Asniéres sur Seine, France) was used. A 3-month follow-up period was applied for those patients in whom DVT was initially excluded. The prevalence of DVT was 18.5% (13/70); four of the diagnoses were made during the 3-month follow-up period. The sensitivity, specificity and negative predictive value of D-dimer were 69.2% (95 confidence interval, 42.4-87.3), 47.4% (95% confidence interval, 35.0-60.1) and 87.1% (95% confidence interval, 71.1-94.9), respectively. In conclusion, D-dimer is of limited value in outpatients on OAT presenting with clinically suspected DVT and should be omitted in such individuals; these patients should always undergo compression venous ultrasound, and repeat ultrasonography within 1 week might be warranted in cases with an initial negative examination.     

Small proportions: what to report for confidence intervals?

The original article to which this Erratum refers was published in Pharmacoepidemiology and Drug Safety 2005; 14: 239–247.     

Low HDL‐cholesterol is common in European Type 2 diabetic patients receiving treatment for dyslipidaemia: data from a pan‐European survey

Aims To measure the prevalence of low high‐density lipoprotein (HDL)‐cholesterol (men < 1.03 mmol/l; women < 1.29 mmol/l) in European Type 2 diabetic patients receiving treatment for dyslipidaemia. Methods The pan‐European Survey of HDL‐cholesterol measured lipids and other cardiovascular risk factors in 3866 patients with Type 2 diabetes and 4436 non‐diabetic patients undergoing treatment for dyslipidaemia in 11 European countries. Results Diabetic patients were more likely to be obese or hypertensive than non‐diabetic patients. Most patients received lifestyle interventions (87%) and/or a statin (89%); treatment patterns were similar between groups. Diabetic patients had [means (SD)] lower HDL‐cholesterol [1.22 (0.37) vs. 1.35 mmol/l (0.44) vs. non‐diabetic patients, P < 0.001] and higher triglycerides [2.32 (2.10) vs. 1.85 mmol/l (1.60), P < 0.001]. More diabetic vs. non‐diabetic patients had low HDL‐cholesterol (45% vs. 30%), high triglycerides (≥ 1.7 mmol/l; 57% vs. 42%) or both (32% vs. 19%). HDL‐cholesterol < 0.9 mmol/l was observed in 18% of diabetic and 12% of non‐diabetic subjects. Differences between diabetic and non‐diabetic groups were slightly greater for women. LDL‐ and total cholesterol were lower in the diabetic group [3.02 (1.05) vs. 3.30 mmol/l (1.14) and 5.12 (1.32) vs. 5.38 mmol/l (1.34), respectively, P < 0.001 for each]. Conclusions Low HDL‐cholesterol is common in diabetes: one in two diabetic women has low HDL‐cholesterol and one diabetic man in four has very low HDL‐cholesterol. Management strategies should include correction of low HDL‐cholesterol to optimize cardiovascular risk in diabetes.