Angela M. Wood

Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls

Most studies have some missing data. Jonathan Sterne and colleagues describe the appropriate use and reporting of the multiple imputation approach to dealing with them

Multiple imputation using chained equations: Issues and guidance for practice

Multiple imputation by chained equations is a flexible and practical approach to handling missing data. We describe the principles of the method and show how to impute categorical and quantitative variables, including skewed variables. We give guidance on how to specify the imputation model and how many imputations are needed. We describe the practical analysis of multiply imputed data, including model building and model checking. We stress the limitations of the method and discuss the possible pitfalls. We illustrate the ideas using a data set in mental health, giving Stata code fragments.

Major lipids, apolipoproteins, and risk of vascular disease.

CONTEXT Associations of major lipids and apolipoproteins with the risk of vascular disease have not been reliably quantified. OBJECTIVE To assess major lipids and apolipoproteins in vascular risk. DESIGN, SETTING, AND PARTICIPANTS Individual records were supplied on 302,430 people without initial vascular disease from 68 long-term prospective studies, mostly in Europe and North America. During 2.79 million person-years of follow-up, there were 8857 nonfatal myocardial infarctions, 3928 coronary heart disease [CHD] deaths, 2534 ischemic strokes, 513 hemorrhagic strokes, and 2536 unclassified strokes. MAIN OUTCOME MEASURES Hazard ratios (HRs), adjusted for several conventional factors, were calculated for 1-SD higher values: 0.52 log(e) triglyceride, 15 mg/dL high-density lipoprotein cholesterol (HDL-C), 43 mg/dL non-HDL-C, 29 mg/dL apolipoprotein AI, 29 mg/dL apolipoprotein B, and 33 mg/dL directly measured low-density lipoprotein cholesterol (LDL-C). Within-study regression analyses were adjusted for within-person variation and combined using meta-analysis. RESULTS The rates of CHD per 1000 person-years in the bottom and top thirds of baseline lipid distributions, respectively, were 2.6 and 6.2 with triglyceride, 6.4 and 2.4 with HDL-C, and 2.3 and 6.7 with non-HDL-C. Adjusted HRs for CHD were 0.99 (95% CI, 0.94-1.05) with triglyceride, 0.78 (95% CI, 0.74-0.82) with HDL-C, and 1.50 (95% CI, 1.39-1.61) with non-HDL-C. Hazard ratios were at least as strong in participants who did not fast as in those who did. The HR for CHD was 0.35 (95% CI, 0.30-0.42) with a combination of 80 mg/dL lower non-HDL-C and 15 mg/dL higher HDL-C. For the subset with apolipoproteins or directly measured LDL-C, HRs were 1.50 (95% CI, 1.38-1.62) with the ratio non-HDL-C/HDL-C, 1.49 (95% CI, 1.39-1.60) with the ratio apo B/apo AI, 1.42 (95% CI, 1.06-1.91) with non-HDL-C, and 1.38 (95% CI, 1.09-1.73) with directly measured LDL-C. Hazard ratios for ischemic stroke were 1.02 (95% CI, 0.94-1.11) with triglyceride, 0.93 (95% CI, 0.84-1.02) with HDL-C, and 1.12 (95% CI, 1.04-1.20) with non-HDL-C. CONCLUSION Lipid assessment in vascular disease can be simplified by measurement of either total and HDL cholesterol levels or apolipoproteins without the need to fast and without regard to triglyceride.

Separate and combined associations of body-mass index and abdominal adiposity with cardiovascular disease : collaborative analysis of 58 prospective studies

BACKGROUND Guidelines differ about the value of assessment of adiposity measures for cardiovascular disease risk prediction when information is available for other risk factors. We studied the separate and combined associations of body-mass index (BMI), waist circumference, and waist-to-hip ratio with risk of first-onset cardiovascular disease. METHODS We used individual records from 58 cohorts to calculate hazard ratios (HRs) per 1 SD higher baseline values (4.56 kg/m(2) higher BMI, 12.6 cm higher waist circumference, and 0.083 higher waist-to-hip ratio) and measures of risk discrimination and reclassification. Serial adiposity assessments were used to calculate regression dilution ratios. RESULTS Individual records were available for 221,934 people in 17 countries (14,297 incident cardiovascular disease outcomes; 1.87 million person-years at risk). Serial adiposity assessments were made in up to 63,821 people (mean interval 5.7 years [SD 3.9]). In people with BMI of 20 kg/m(2) or higher, HRs for cardiovascular disease were 1.23 (95% CI 1.17-1.29) with BMI, 1.27 (1.20-1.33) with waist circumference, and 1.25 (1.19-1.31) with waist-to-hip ratio, after adjustment for age, sex, and smoking status. After further adjustment for baseline systolic blood pressure, history of diabetes, and total and HDL cholesterol, corresponding HRs were 1.07 (1.03-1.11) with BMI, 1.10 (1.05-1.14) with waist circumference, and 1.12 (1.08-1.15) with waist-to-hip ratio. Addition of information on BMI, waist circumference, or waist-to-hip ratio to a cardiovascular disease risk prediction model containing conventional risk factors did not importantly improve risk discrimination (C-index changes of -0.0001, -0.0001, and 0.0008, respectively), nor classification of participants to categories of predicted 10-year risk (net reclassification improvement -0.19%, -0.05%, and -0.05%, respectively). Findings were similar when adiposity measures were considered in combination. Reproducibility was greater for BMI (regression dilution ratio 0.95, 95% CI 0.93-0.97) than for waist circumference (0.86, 0.83-0.89) or waist-to-hip ratio (0.63, 0.57-0.70). INTERPRETATION BMI, waist circumference, and waist-to-hip ratio, whether assessed singly or in combination, do not importantly improve cardiovascular disease risk prediction in people in developed countries when additional information is available for systolic blood pressure, history of diabetes, and lipids. FUNDING British Heart Foundation and UK Medical Research Council.Summary Background Guidelines differ about the value of assessment of adiposity measures for cardiovascular disease risk prediction when information is available for other risk factors. We studied the separate and combined associations of body-mass index (BMI), waist circumference, and waist-to-hip ratio with risk of first-onset cardiovascular disease. Methods We used individual records from 58 cohorts to calculate hazard ratios (HRs) per 1 SD higher baseline values (4·56 kg/m2 higher BMI, 12·6 cm higher waist circumference, and 0·083 higher waist-to-hip ratio) and measures of risk discrimination and reclassification. Serial adiposity assessments were used to calculate regression dilution ratios. Results Individual records were available for 221 934 people in 17 countries (14 297 incident cardiovascular disease outcomes; 1·87 million person-years at risk). Serial adiposity assessments were made in up to 63 821 people (mean interval 5·7 years [SD 3·9]). In people with BMI of 20 kg/m2 or higher, HRs for cardiovascular disease were 1·23 (95% CI 1·17–1·29) with BMI, 1·27 (1·20–1·33) with waist circumference, and 1·25 (1·19–1·31) with waist-to-hip ratio, after adjustment for age, sex, and smoking status. After further adjustment for baseline systolic blood pressure, history of diabetes, and total and HDL cholesterol, corresponding HRs were 1·07 (1·03–1·11) with BMI, 1·10 (1·05–1·14) with waist circumference, and 1·12 (1·08–1·15) with waist-to-hip ratio. Addition of information on BMI, waist circumference, or waist-to-hip ratio to a cardiovascular disease risk prediction model containing conventional risk factors did not importantly improve risk discrimination (C-index changes of −0·0001, −0·0001, and 0·0008, respectively), nor classification of participants to categories of predicted 10-year risk (net reclassification improvement −0·19%, −0·05%, and −0·05%, respectively). Findings were similar when adiposity measures were considered in combination. Reproducibility was greater for BMI (regression dilution ratio 0·95, 95% CI 0·93–0·97) than for waist circumference (0·86, 0·83–0·89) or waist-to-hip ratio (0·63, 0·57–0·70). Interpretation BMI, waist circumference, and waist-to-hip ratio, whether assessed singly or in combination, do not importantly improve cardiovascular disease risk prediction in people in developed countries when additional information is available for systolic blood pressure, history of diabetes, and lipids. Funding British Heart Foundation and UK Medical Research Council.

Long-Term Interleukin-6 Levels and Subsequent Risk of Coronary Heart Disease: Two New Prospective Studies and a Systematic Review

Background The relevance to coronary heart disease (CHD) of cytokines that govern inflammatory cascades, such as interleukin-6 (IL-6), may be underestimated because such mediators are short acting and prone to fluctuations. We evaluated associations of long-term circulating IL-6 levels with CHD risk (defined as nonfatal myocardial infarction [MI] or fatal CHD) in two population-based cohorts, involving serial measurements to enable correction for within-person variability. We updated a systematic review to put the new findings in context. Methods and Findings Measurements were made in samples obtained at baseline from 2,138 patients who had a first-ever nonfatal MI or died of CHD during follow-up, and from 4,267 controls in two cohorts comprising 24,230 participants. Correction for within-person variability was made using data from repeat measurements taken several years apart in several hundred participants. The year-to-year variability of IL-6 values within individuals was relatively high (regression dilution ratios of 0.41, 95% confidence interval [CI] 0.28–0.53, over 4 y, and 0.35, 95% CI 0.23–0.48, over 12 y). Ignoring this variability, we found an odds ratio for CHD, adjusted for several established risk factors, of 1.46 (95% CI 1.29–1.65) per 2 standard deviation (SD) increase of baseline IL-6 values, similar to that for baseline C-reactive protein. After correction for within-person variability, the odds ratio for CHD was 2.14 (95% CI 1.45–3.15) with long-term average (“usual”) IL-6, similar to those for some established risk factors. Increasing IL-6 levels were associated with progressively increasing CHD risk. An updated systematic review of electronic databases and other sources identified 15 relevant previous population-based prospective studies of IL-6 and clinical coronary outcomes (i.e., MI or coronary death). Including the two current studies, the 17 available prospective studies gave a combined odds ratio of 1.61 (95% CI 1.42–1.83) per 2 SD increase in baseline IL-6 (corresponding to an odds ratio of 3.34 [95% CI 2.45–4.56] per 2 SD increase in usual [long-term average] IL-6 levels). Conclusions Long-term IL-6 levels are associated with CHD risk about as strongly as are some major established risk factors, but causality remains uncertain. These findings highlight the potential relevance of IL-6–mediated pathways to CHD.

Separate and combined associations of body-mass index and abdominal adiposity with cardiovascular disease: collaborative analysis of 58 prospective studies

BACKGROUND Guidelines differ about the value of assessment of adiposity measures for cardiovascular disease risk prediction when information is available for other risk factors. We studied the separate and combined associations of body-mass index (BMI), waist circumference, and waist-to-hip ratio with risk of first-onset cardiovascular disease. METHODS We used individual records from 58 cohorts to calculate hazard ratios (HRs) per 1 SD higher baseline values (4.56 kg/m(2) higher BMI, 12.6 cm higher waist circumference, and 0.083 higher waist-to-hip ratio) and measures of risk discrimination and reclassification. Serial adiposity assessments were used to calculate regression dilution ratios. RESULTS Individual records were available for 221,934 people in 17 countries (14,297 incident cardiovascular disease outcomes; 1.87 million person-years at risk). Serial adiposity assessments were made in up to 63,821 people (mean interval 5.7 years [SD 3.9]). In people with BMI of 20 kg/m(2) or higher, HRs for cardiovascular disease were 1.23 (95% CI 1.17-1.29) with BMI, 1.27 (1.20-1.33) with waist circumference, and 1.25 (1.19-1.31) with waist-to-hip ratio, after adjustment for age, sex, and smoking status. After further adjustment for baseline systolic blood pressure, history of diabetes, and total and HDL cholesterol, corresponding HRs were 1.07 (1.03-1.11) with BMI, 1.10 (1.05-1.14) with waist circumference, and 1.12 (1.08-1.15) with waist-to-hip ratio. Addition of information on BMI, waist circumference, or waist-to-hip ratio to a cardiovascular disease risk prediction model containing conventional risk factors did not importantly improve risk discrimination (C-index changes of -0.0001, -0.0001, and 0.0008, respectively), nor classification of participants to categories of predicted 10-year risk (net reclassification improvement -0.19%, -0.05%, and -0.05%, respectively). Findings were similar when adiposity measures were considered in combination. Reproducibility was greater for BMI (regression dilution ratio 0.95, 95% CI 0.93-0.97) than for waist circumference (0.86, 0.83-0.89) or waist-to-hip ratio (0.63, 0.57-0.70). INTERPRETATION BMI, waist circumference, and waist-to-hip ratio, whether assessed singly or in combination, do not importantly improve cardiovascular disease risk prediction in people in developed countries when additional information is available for systolic blood pressure, history of diabetes, and lipids. FUNDING British Heart Foundation and UK Medical Research Council.Summary Background Guidelines differ about the value of assessment of adiposity measures for cardiovascular disease risk prediction when information is available for other risk factors. We studied the separate and combined associations of body-mass index (BMI), waist circumference, and waist-to-hip ratio with risk of first-onset cardiovascular disease. Methods We used individual records from 58 cohorts to calculate hazard ratios (HRs) per 1 SD higher baseline values (4·56 kg/m2 higher BMI, 12·6 cm higher waist circumference, and 0·083 higher waist-to-hip ratio) and measures of risk discrimination and reclassification. Serial adiposity assessments were used to calculate regression dilution ratios. Results Individual records were available for 221 934 people in 17 countries (14 297 incident cardiovascular disease outcomes; 1·87 million person-years at risk). Serial adiposity assessments were made in up to 63 821 people (mean interval 5·7 years [SD 3·9]). In people with BMI of 20 kg/m2 or higher, HRs for cardiovascular disease were 1·23 (95% CI 1·17–1·29) with BMI, 1·27 (1·20–1·33) with waist circumference, and 1·25 (1·19–1·31) with waist-to-hip ratio, after adjustment for age, sex, and smoking status. After further adjustment for baseline systolic blood pressure, history of diabetes, and total and HDL cholesterol, corresponding HRs were 1·07 (1·03–1·11) with BMI, 1·10 (1·05–1·14) with waist circumference, and 1·12 (1·08–1·15) with waist-to-hip ratio. Addition of information on BMI, waist circumference, or waist-to-hip ratio to a cardiovascular disease risk prediction model containing conventional risk factors did not importantly improve risk discrimination (C-index changes of −0·0001, −0·0001, and 0·0008, respectively), nor classification of participants to categories of predicted 10-year risk (net reclassification improvement −0·19%, −0·05%, and −0·05%, respectively). Findings were similar when adiposity measures were considered in combination. Reproducibility was greater for BMI (regression dilution ratio 0·95, 95% CI 0·93–0·97) than for waist circumference (0·86, 0·83–0·89) or waist-to-hip ratio (0·63, 0·57–0·70). Interpretation BMI, waist circumference, and waist-to-hip ratio, whether assessed singly or in combination, do not importantly improve cardiovascular disease risk prediction in people in developed countries when additional information is available for systolic blood pressure, history of diabetes, and lipids. Funding British Heart Foundation and UK Medical Research Council.

Are missing outcome data adequately handled? A review of published randomized controlled trials in major medical journals

Background Randomized controlled trials almost always have some individuals with missing outcomes. Inadequate handling of these missing data in the analysis can cause substantial bias in the treatment effect estimates. We examine how missing outcome data are handled in randomized controlled trials in order to assess whether adequate steps have been taken to reduce nonresponse bias and to identify ways to improve procedures for missing data. Methods We reviewed all randomized trials published between July and December 2001 in BMJ, JAMA, Lancet and New England Journal of Medicine, excluding trials in which the primary outcome was described as a time-to-event. We focused on trial designs, how missing outcome data were described and the statistical methods used to deal with the missing outcome data, including sensitivity analyses. Results We identified 71 trials of which 63 (89%) reported having partly missing outcome data: 13 trials had more than 20% of patients with missing outcomes. In 26 trials that measured the outcome at a single time point, 92% performed a complete case analysis and 8% imputed the missing outcomes using baseline values or the worst case value. In 37 trials with repeated measures of the outcome, 46% performed complete case analyses, potentially excluding individuals with some follow-up data, while 14% performed a repeated measures analysis, 19% used the last observation carried forward, 11% imputed with the worst case value and 2% imputed using regression predictions. Thirteen (21%) of trials with missing data reported a sensitivity analysis. Conclusions Our review shows that missing outcome data are a common problem in randomized controlled trials, and are often inadequately handled in the statistical analysis in the top tier medical journals. Authors should explicitly state the assumptions underlying the handling of the missing outcomes and justify them through data descriptions and sensitivity analyses.

Lipid-related markers and cardiovascular disease prediction.

CONTEXT The value of assessing various emerging lipid-related markers for prediction of first cardiovascular events is debated. OBJECTIVE To determine whether adding information on apolipoprotein B and apolipoprotein A-I, lipoprotein(a), or lipoprotein-associated phospholipase A2 to total cholesterol and high-density lipoprotein cholesterol (HDL-C) improves cardiovascular disease (CVD) risk prediction. DESIGN, SETTING, AND PARTICIPANTS Individual records were available for 165,544 participants without baseline CVD in 37 prospective cohorts (calendar years of recruitment: 1968-2007) with up to 15,126 incident fatal or nonfatal CVD outcomes (10,132 CHD and 4994 stroke outcomes) during a median follow-up of 10.4 years (interquartile range, 7.6-14 years). MAIN OUTCOME MEASURES Discrimination of CVD outcomes and reclassification of participants across predicted 10-year risk categories of low (<10%), intermediate (10%-<20%), and high (≥20%) risk. RESULTS The addition of information on various lipid-related markers to total cholesterol, HDL-C, and other conventional risk factors yielded improvement in the models discrimination: C-index change, 0.0006 (95% CI, 0.0002-0.0009) for the combination of apolipoprotein B and A-I; 0.0016 (95% CI, 0.0009-0.0023) for lipoprotein(a); and 0.0018 (95% CI, 0.0010-0.0026) for lipoprotein-associated phospholipase A2 mass. Net reclassification improvements were less than 1% with the addition of each of these markers to risk scores containing conventional risk factors. We estimated that for 100,000 adults aged 40 years or older, 15,436 would be initially classified at intermediate risk using conventional risk factors alone. Additional testing with a combination of apolipoprotein B and A-I would reclassify 1.1%; lipoprotein(a), 4.1%; and lipoprotein-associated phospholipase A2 mass, 2.7% of people to a 20% or higher predicted CVD risk category and, therefore, in need of statin treatment under Adult Treatment Panel III guidelines. CONCLUSION In a study of individuals without known CVD, the addition of information on the combination of apolipoprotein B and A-I, lipoprotein(a), or lipoprotein-associated phospholipase A2 mass to risk scores containing total cholesterol and HDL-C led to slight improvement in CVD prediction.

Chocolate consumption and cardiometabolic disorders: systematic review and meta-analysis

Objective To evaluate the association of chocolate consumption with the risk of developing cardiometabolic disorders. Design Systematic review and meta-analysis of randomised controlled trials and observational studies. Data sources Medline, Embase, Cochrane Library, PubMed, CINAHL, IPA, Web of Science, Scopus, Pascal, reference lists of relevant studies to October 2010, and email contact with authors. Study selection Randomised trials and cohort, case-control, and cross sectional studies carried out in human adults, in which the association between chocolate consumption and the risk of outcomes related to cardiometabolic disorders were reported. Data extraction Data were extracted by two independent investigators, and a consensus was reached with the involvement of a third. The primary outcome was cardiometabolic disorders, including cardiovascular disease (coronary heart disease and stroke), diabetes, and metabolic syndrome. A meta-analysis assessed the risk of developing cardiometabolic disorders by comparing the highest and lowest level of chocolate consumption. Results From 4576 references seven studies met the inclusion criteria (including 114 009 participants). None of the studies was a randomised trial, six were cohort studies, and one a cross sectional study. Large variation was observed between these seven studies for measurement of chocolate consumption, methods, and outcomes evaluated. Five of the seven studies reported a beneficial association between higher levels of chocolate consumption and the risk of cardiometabolic disorders. The highest levels of chocolate consumption were associated with a 37% reduction in cardiovascular disease (relative risk 0.63 (95% confidence interval 0.44 to 0.90)) and a 29% reduction in stroke compared with the lowest levels. Conclusions Based on observational evidence, levels of chocolate consumption seem to be associated with a substantial reduction in the risk of cardiometabolic disorders. Further experimental studies are required to confirm a potentially beneficial effect of chocolate consumption.

Association of Cardiometabolic Multimorbidity With Mortality.

IMPORTANCE The prevalence of cardiometabolic multimorbidity is increasing. OBJECTIVE To estimate reductions in life expectancy associated with cardiometabolic multimorbidity. DESIGN, SETTING, AND PARTICIPANTS Age- and sex-adjusted mortality rates and hazard ratios (HRs) were calculated using individual participant data from the Emerging Risk Factors Collaboration (689,300 participants; 91 cohorts; years of baseline surveys: 1960-2007; latest mortality follow-up: April 2013; 128,843 deaths). The HRs from the Emerging Risk Factors Collaboration were compared with those from the UK Biobank (499,808 participants; years of baseline surveys: 2006-2010; latest mortality follow-up: November 2013; 7995 deaths). Cumulative survival was estimated by applying calculated age-specific HRs for mortality to contemporary US age-specific death rates. EXPOSURES A history of 2 or more of the following: diabetes mellitus, stroke, myocardial infarction (MI). MAIN OUTCOMES AND MEASURES All-cause mortality and estimated reductions in life expectancy. RESULTS In participants in the Emerging Risk Factors Collaboration without a history of diabetes, stroke, or MI at baseline (reference group), the all-cause mortality rate adjusted to the age of 60 years was 6.8 per 1000 person-years. Mortality rates per 1000 person-years were 15.6 in participants with a history of diabetes, 16.1 in those with stroke, 16.8 in those with MI, 32.0 in those with both diabetes and MI, 32.5 in those with both diabetes and stroke, 32.8 in those with both stroke and MI, and 59.5 in those with diabetes, stroke, and MI. Compared with the reference group, the HRs for all-cause mortality were 1.9 (95% CI, 1.8-2.0) in participants with a history of diabetes, 2.1 (95% CI, 2.0-2.2) in those with stroke, 2.0 (95% CI, 1.9-2.2) in those with MI, 3.7 (95% CI, 3.3-4.1) in those with both diabetes and MI, 3.8 (95% CI, 3.5-4.2) in those with both diabetes and stroke, 3.5 (95% CI, 3.1-4.0) in those with both stroke and MI, and 6.9 (95% CI, 5.7-8.3) in those with diabetes, stroke, and MI. The HRs from the Emerging Risk Factors Collaboration were similar to those from the more recently recruited UK Biobank. The HRs were little changed after further adjustment for markers of established intermediate pathways (eg, levels of lipids and blood pressure) and lifestyle factors (eg, smoking, diet). At the age of 60 years, a history of any 2 of these conditions was associated with 12 years of reduced life expectancy and a history of all 3 of these conditions was associated with 15 years of reduced life expectancy. CONCLUSIONS AND RELEVANCE Mortality associated with a history of diabetes, stroke, or MI was similar for each condition. Because any combination of these conditions was associated with multiplicative mortality risk, life expectancy was substantially lower in people with multimorbidity.