Estimating the contribution of overweight and obesity to ethnic inequalities in cardio-metabolic diseases in the Netherlands: a simulation study.

OBJECTIVES: Overweight and obesity (OWOB) starts in childhood, influences adult cardiovascular risk, and is not equally distributed across ethnic groups. It is unclear which effects can be expected from reductions in OWOB across the life course on inequalities in cardio-metabolic diseases in a multi-ethnic population. This study aims to estimate the effects of three scenarios of changes in OWOB (the Normal-Weight-for-All scenario, the No-Ethnic-Difference-over-the-Life-Course scenario, the and No-Ethnic-Differences-in-Childhood scenario). STUDY DESIGN: A simulation study. METHODS: We combine data from multiple data sources and use the Dynamic Modeling for Health Impact Assessment (DYNAMO-HIA) model to estimate the effects of three scenarios on the cumulative incidence of diabetes mellitus, ischaemic heart disease (IHD) and stroke between 18 and 70 years in the five largest ethnic groups in the Netherlands. RESULTS: In the scenario where all individuals have normal weight, the cumulative incidence decreased in all ethnic minority groups for all diseases, with largest decreases among South-Asian Surinamese, where the reduction of diabetes incidence exceeded 50%. In the scenario where the prevalence of OWOB in each ethnic-minority group was reduced to the current level among the Dutch-origin population, ethnic inequalities in cardio-metabolic diseases were substantially reduced, particularly when lowered prevalence of OWOB persisted across the lifespan. Reductions were the largest for diabetes and for the Asian Surinamese population. CONCLUSIONS: A substantial part of the well-known ethnic inequalities in incidence of diabetes, IHD, and stroke can be attributed to OWOB. Interventions aimed at reducing OWOB have clear potential to reduce the health inequalities in these outcomes, especially for diabetes, in particular when they have an impact across the lifespan.

Mortality in patients with rheumatoid arthritis: a 15-year prospective cohort study.

The aim of this study was to investigate (a) the mortality in a clinical cohort of patients with established rheumatoid arthritis in comparison with the general Dutch population over 15 years, (b) the trend in the mortality ratio during the study period, and (c) causes of death and compare these with the general population. In 1997, a sample of 1222 patients was randomly selected from the register of a large rheumatology outpatient clinic. Their mortality and primary causes of death between 1997 and 2012 were obtained from Statistics Netherlands. The standardized mortality ratio (SMR) for all-cause mortality and the number of life-years lost in the study period, adjusted for age, sex, and calendar year, were calculated. A linear poisson regression analysis was performed to evaluate change in all-cause SMR over time. Finally, the SMRs for cause-specific mortality were calculated. The mean age of the population at baseline was 60.4 (SD 15.4) years, and 72.6% of the patients were women. The estimated SMR (95% CI) for all-cause mortality was 1.54 (1.41, 1.67) with about one life-year lost over the study period. There was a trend to decreasing SMR (2% annually, p = .07). Mortality was higher compared with the general population for circulatory system diseases, respiratory system diseases, musculoskeletal system diseases, and digestive system diseases (p < .05). The observed mortality among patients with RA was 54% higher than in the general population after adjustment for age, sex and calendar year. More than one life-year was lost over 15 years, and the mortality tended to decrease over time. The mortality was higher for cardiovascular, respiratory, musculoskeletal and digestive diseases.

Predicting the health impact of lowering salt consumption in Turkey using the DYNAMO health impact assessment tool.

OBJECTIVE: To estimate the impact of three daily salt consumption scenarios on the prevalence and incidence of ischaemic heart disease (IHD) and cerebrovascular disease in 2025 in the Turkish population aged ≥30 years using the DYNAMO Health Impact Assessment tool. STUDY DESIGN: Statistical disease modelling study. METHODS: DYNAMO health impact assessment was populated using data from Turkey to estimate the prevalence and incidence of IHD and cerebrovascular disease in 2025. TurkSTAT data were used for demographic data, and national surveys were used for salt consumption and disease-specific burden. Three salt consumption scenarios were modelled: (1) reference scenario: mean salt consumption stays the same from 2012-2013 until 2025; (2) gradual decline: daily salt intake reduces steadily by 0.47 g per year by lowering salt intake from bread by 50% and from table salt by 40% by 2025; and (3) World Health Organization (WHO) advice: daily salt intake of 5 g per day from 2013 until 2025. RESULTS: The gradual decline scenario would lead to a decrease in the prevalence of IHD and cerebrovascular disease by 0.3% and 0.2%, respectively, and a decrease in the incidence by 0.6 and 0.4 per 1000, respectively. Following WHO's advice would lead to a decrease in the prevalence of IHD and cerebrovascular disease by 0.8% and 0.5%, respectively, and a decrease in the incidence by 1.0 and 0.7 per 1000, respectively. CONCLUSION: This model indicates that Turkey can lower its future cardiovascular disease burden by following the gradual decline scenario. Following WHO's advice would achieve an even greater benefit.

Screen time, adiposity and cardiometabolic markers: mediation by physical activity, not snacking, among 11-year-old children.

BACKGROUND: There is evidence for a relation of TV viewing with adiposity and increased cardiometabolic risk factors in children and adolescents. It is unclear to what extent this relation is mediated by snacking and lack of physical activity. We determined whether associations of screen time with adiposity and cardiometabolic markers were mediated by these behaviours. METHODS: Children from a population-representative Dutch birth cohort (n=1447) reported screen time and other lifestyle factors by a questionnaire around the age of 11 years (range 10-14) and had anthropometry and cardiometabolic markers measured around the age of 12 years (range 12-14). Adjusted associations of screen time with snacking, physical activity, adiposity and cardiometabolic markers (total-to-high-density lipoprotein cholesterol (TC/HDLC) ratio, blood pressure, glycated haemoglobin) were assessed by using formal mediation analysis. We tested the hypothesized paths by structural equation modeling, which allows quantification of the indirect effects associated with potential mediators. RESULTS: Children with ⩾20 h screen time per week consumed more snacks (1.9 vs 1.3 portions per day) and were less physically active (4.3 vs 4.8 days per week) than children with maximum 6 h screen time. Screen time was directly associated with higher adiposity (standardized ß=0.10-0.12 depending on the outcome, P<0.001), and indirectly through less physical activity. The association of screen time with TC/HDLC ratio was almost completely mediated by adiposity (ß=0.39, P<0.0001), and to a minor extent by physical activity (ß=-0.06, P=0.02). There was no direct association of screen time with TC/HDLC ratio. CONCLUSIONS: The adverse association of screen time with adiposity was partly mediated by physical activity, but not by snacking. The association of screen time with TC/HDLC ratio was almost completely mediated by adiposity. Our results may suggest that future efforts in society and public health should be directed to replace screen time with physical activity for reducing children's adiposity and cardiometabolic risk.

Variety in vegetable and fruit consumption and the risk of gastric and esophageal cancer in the European Prospective Investigation into Cancer and Nutrition.

Diets high in vegetables and fruits have been suggested to be inversely associated with risk of gastric cancer. However, the evidence of the effect of variety of consumption is limited. We therefore investigated whether consumption of a variety of vegetables and fruit is associated with gastric and esophageal cancer in the European Prospective Investigation into Cancer and Nutrition study. Data on food consumption and follow-up on cancer incidence were available for 452,269 participants from 10 European countries. After a mean follow-up of 8.4 years, 475 cases of gastric and esophageal adenocarcinomas (180 noncardia, 185 cardia, gastric esophageal junction and esophagus, 110 not specified) and 98 esophageal squamous cell carcinomas were observed. Diet Diversity Scores were used to quantify the variety in vegetable and fruit consumption. We used multivariable Cox proportional hazard models to calculate risk ratios. Independent from quantity of consumption, variety in the consumption of vegetables and fruit combined and of fruit consumption alone were statistically significantly inversely associated with the risk of esophageal squamous cell carcinoma (continuous hazard ratio per 2 products increment 0.88; 95% CI 0.79-0.97 and 0.76; 95% CI 0.62-0.94, respectively) with the latter particularly seen in ever smokers. Variety in vegetable and/or fruit consumption was not associated with risk of gastric and esophageal adenocarcinomas. Independent from quantity of consumption, more variety in vegetable and fruit consumption combined and in fruit consumption alone may decrease the risk of esophageal squamous cell carcinoma. However, residual confounding by lifestyle factors cannot be excluded.

Estimating net transition probabilities from cross-sectional data with application to risk factors in chronic disease modeling.

A problem occurring in chronic disease modeling is the estimation of transition probabilities of moving from one state of a categorical risk factor to another. Transitions could be obtained from a cohort study, but often such data may not be available. However, under the assumption that transitions remain stable over time, age specific cross-sectional prevalence data could be used instead. Problems that then arise are parameter identifiability and the fact that age dependent cross-sectional data are often noisy or are given in age intervals. In this paper we propose a method to estimate so-called net annual transition probabilities from cross-sectional data, including their uncertainties. Net transitions only describe the net inflow or outflow into a certain risk factor state at a certain age. Our approach consists of two steps: first, smooth the data using multinomial P-splines, second, from these data estimate net transition probabilities. This second step can be formulated as a transportation problem, which is solved using the simplex algorithm from linear programming theory. A sensible specification of the cost matrix is crucial to get meaningful results. Uncertainties are assessed by parametric bootstrapping. We illustrate our method using data on body mass index. We conclude that this method provides a flexible way of estimating net transitions and that the use of net transitions has implications for model dynamics, for example when modeling interventions.

Using random forest to identify longitudinal predictors of health in a 30-year cohort study.

Due to the wealth of exposome data from longitudinal cohort studies that is currently available, the need for methods to adequately analyze these data is growing. We propose an approach in which machine learning is used to identify longitudinal exposome-related predictors of health, and illustrate its potential through an application. Our application involves studying the relation between exposome and self-perceived health based on the 30-year running Doetinchem Cohort Study. Random Forest (RF) was used to identify the strongest predictors due to its favorable prediction performance in prior research. The relation between predictors and outcome was visualized with partial dependence and accumulated local effects plots. To facilitate interpretation, exposures were summarized by expressing them as the average exposure and average trend over time. The RF model's ability to discriminate poor from good self-perceived health was acceptable (Area-Under-the-Curve = 0.707). Nine exposures from different exposome-related domains were largely responsible for the model's performance, while 87 exposures seemed to contribute little to the performance. Our approach demonstrates that ML can be interpreted more than widely believed, and can be applied to identify important longitudinal predictors of health over the life course in studies with repeated measures of exposure. The approach is context-independent and broadly applicable.

The influence of population characteristics on variation in general practice based morbidity estimations.

BACKGROUND: General practice based registration networks (GPRNs) provide information on morbidity rates in the population. Morbidity rate estimates from different GPRNs, however, reveal considerable, unexplained differences. We studied the range and variation in morbidity estimates, as well as the extent to which the differences in morbidity rates between general practices and networks change if socio-demographic characteristics of the listed patient populations are taken into account. METHODS: The variation in incidence and prevalence rates of thirteen diseases among six Dutch GPRNs and the influence of age, gender, socio economic status (SES), urbanization level, and ethnicity are analyzed using multilevel logistic regression analysis. Results are expressed in median odds ratios (MOR). RESULTS: We observed large differences in morbidity rate estimates both on the level of general practices as on the level of networks. The differences in SES, urbanization level and ethnicity distribution among the networks' practice populations are substantial. The variation in morbidity rate estimates among networks did not decrease after adjusting for these socio-demographic characteristics. CONCLUSION: Socio-demographic characteristics of populations do not explain the differences in morbidity estimations among GPRNs.

Fruits and vegetables consumption and the risk of histological subtypes of lung cancer in the European Prospective Investigation into Cancer and Nutrition (EPIC).

OBJECTIVE: To examine the association between fruit and vegetable consumption and risk of different histological subtypes of lung cancer among participants of the European Prospective Investigation into Cancer and Nutrition study. METHODS: Multivariable Cox proportional hazard models were used to analyze the data. A calibration study in a subsample was used to reduce dietary measurement errors. RESULTS: During a mean follow-up of 8.7 years, 1,830 incident cases of lung cancer (574 adenocarcinoma, 286 small cell, 137 large cell, 363 squamous cell, 470 other histologies) were identified. In line with our previous conclusions, we found that after calibration a 100 g/day increase in fruit and vegetables consumption was associated with a reduced lung cancer risk (HR 0.94; 95% CI 0.89-0.99). This was also seen among current smokers (HR 0.93; 95% CI 0.90-0.97). Risks of squamous cell carcinomas in current smokers were reduced for an increase of 100 g/day of fruit and vegetables combined (HR 0.85; 95% CI 0.76-0.94), while no clear effects were seen for the other histological subtypes. CONCLUSION: We observed inverse associations between the consumption of vegetables and fruits and risk of lung cancer without a clear effect on specific histological subtypes of lung cancer. In current smokers, consumption of vegetables and fruits may reduce lung cancer risk, in particular the risk of squamous cell carcinomas.

High prevalence of methicillin resistant Staphylococcus aureus in pigs.

Recently methicillin resistant Staphylococcus aureus (MRSA) was isolated from pigs and pig farmers in The Netherlands. In order to assess the dissemination of MRSA in the Dutch pig population, we screened 540 pigs in 9 slaughterhouses, where a representative portion of Dutch pigs (63%) was slaughtered in 2005. We found 209 (39%) of the pigs to carry MRSA in their nares. Forty-four of 54 groups of 10 consecutive pigs (81%), each group from a different farm, and all slaughterhouses were affected. All MRSA isolates belonged to 1 clonal group, showing Multi-Locus Sequence Type 398 and closely related spa types (mainly t011, t108 and t1254). Three types of the Staphylococcal Chromosome Cassette (SCCmec) were found: III (3%), IVa (39%) and V (57%). All 44 tested isolates (1 isolate per group) were resistant to tetracycline, reflecting the high and predominant use of tetracyclines in pig husbandry. Twenty-three percent of the isolates were resistant to both erythromycin and clindamycin and 36% to kanamycin, gentamicin and tobramycin but only a single isolate was resistant to co-trimoxazole and none to ciprofloxacin and several other antibiotics. The percentage of MRSA positive pigs was significantly different among slaughterhouses and among groups within slaughterhouses, indicating a high prevalence of MRSA in pigs delivered from the farms as well as cross contamination in the slaughterhouses.

Blood pressure trajectories in relation to cardiovascular mortality: The Rancho Bernardo Study.

The added value of blood pressure (BP) trajectories for predicting cardiovascular disease (CVD) is currently unknown. We investigated the association of systolic BP (SBP) trajectories with CVD and all-cause mortality and compared these associations with those of average SBP, taking antihypertensive medication into account. Data from 762 participants of the Rancho Bernardo Study were used. SBP from five examinations (maximum) from 1984 to 2002 was used; mortality data were obtained from 2002 to 2013. SBP trajectories were derived using group-based trajectory modelling. Cox proportional hazards analysis was used to investigate associations of trajectories and average SBP with CVD and all-cause mortality, adjusted for age, sex, cholesterol, smoking, diabetes and antihypertensive medication. Mean baseline age was 65.7 years, and 67% were women. Four trajectories were identified, in which mean SBP increased by 5-12 mm Hg during 10 years. The highest trajectories were associated with two to three times greater CVD mortality and 1.5 times greater all-cause mortality risk, compared with the lowest trajectory. Each 20 mmHg increment in average SBP was associated with 1.4 times greater CVD mortality risk and 1.2 times all-cause mortality risk. Associations were not modified by antihypertensive medication (P-interaction>0.10). SBP trajectories were not superior to average SBP in predicting CVD and all-cause mortality. In the general middle-aged and older population of the Rancho Bernardo study, SBP trajectories provided no added value to average SBP in predicting CVD and all-cause mortality. Long-term average SBP levels and trajectories were significant predictors of CVD and all-cause mortality, irrespective of prescribed antihypertensive medication (which in the 1980s-1990s mainly were diuretics and ß-blockers).

Effect size estimates of lifestyle and dietary changes on all-cause mortality in coronary artery disease patients: a systematic review.

BACKGROUND: Guidelines for lifestyle and dietary modification in patients with coronary artery disease (CAD) are mainly supported by evidence from general population studies. CAD patients, however, differ from the general population in age (older) and treatment with preventive drugs. This review seeks to provide evidence for a prognostic benefit of lifestyle and dietary recommendations from studies in CAD patients. METHODS AND RESULTS: A literature search was performed on the effect of lifestyle and dietary changes on mortality in CAD patients. Prospective cohort studies and randomized controlled trials of patients with established CAD were included if they reported all-causes mortality and had at least 6 months of follow-up. The effect estimates of smoking cessation (relative risk [RR], 0.64; 95% CI, 0.58 to 0.71), increased physical activity (RR, 0.76; 95% CI, 0.59 to 0.98), and moderate alcohol use (RR, 0.80; 95% CI, 0.78 to 0.83) were studied most extensively. For the 6 dietary goals, data were too limited to provide reliable effect size estimates. Combinations of dietary changes were associated with reduced mortality (RR, 0.56; 95% CI, 0.42 to 0.74). CONCLUSIONS: Available studies show convincingly the health benefits of lifestyle changes in CAD patients. Effect estimates of combined dietary changes look promising. Future studies should confirm these findings and assess the contribution of the individual dietary factors.

Disparities in stroke preventive care in general practice did not explain socioeconomic disparities in stroke.

OBJECTIVE: To assess socioeconomic disparities in stroke incidence and in the quality of preventive care for stroke in the Netherlands. STUDY DESIGN AND SETTINGS: A total of 190,664 patients who registered in 96 general practices were followed up for 12 months. Data were collected on diagnoses, referrals, prescriptions, and diagnostic procedures. Hazard ratios (HR) were calculated to assess the association between educational level and stroke incidence. Multilevel logistic regression was used to assess socioeconomic disparities in the quality of preventive care for stroke precursors. RESULTS: Lower educational level was associated with higher incidence of stroke in men (HR=1.36, 95% CI=1.06-1.74) but not in women. Among both men and women, there were socioeconomic disparities in the prevalence of hypertension, hypercholesterolemia, diabetes, angina pectoris, heart failure, and peripheral artery disease. Lower educated hypercholesterolemia patients under medication were less likely to be prescribed statins (odds ratio=0.62, 95% CI=0.42-0.91). However, for other precursors of stroke, there were no major disparities in the quality of preventive care. CONCLUSION: There are socioeconomic disparities in stroke incidence among men but not among women. Socioeconomic differences in factors such as hypertension and diabetes are likely to contribute to stroke disparities. However, general practitioners (GPs) provide care of a similar quality to patients from different socioeconomic groups.

Health care costs attributable to overweight calculated in a standardized way for three European countries.

This article presents a tool to calculate health care costs attributable to overweight in a comparable and standardized way. The purpose is to describe the methodological principles of the tool and to put it into use by calculating and comparing the costs attributable to overweight for The Netherlands, Germany and Czech Republic. The tool uses a top-down and prevalence-based approach, consisting of five steps. Step one identifies overweight-related diseases and age- and gender-specific relative risks. Included diseases are ischemic heart disease, stroke, hypertension, type 2 diabetes mellitus, colorectal cancer, postmenopausal breast cancer, endometrial cancer, kidney cancer and osteoarthritis. Step two consists of collecting data on the age- and gender-specific prevalence of these diseases. Step three uses the population-attributable prevalence to determine the part of the prevalence of these diseases that is attributable to overweight. Step four calculates the health care costs associated with these diseases. Step five calculates the costs of these diseases that are attributable to overweight. Overweight is responsible for 20-26% of the direct costs of included diseases, with sensitivity analyses varying this percentage between 15-31%. Percentage of costs attributable to obesity and preobesity is about the same. Diseases with the highest percentage of costs due to overweight are diabetes, endometrial cancer and osteoarthritis. Disease costs attributable to overweight as a percentage of total health care expenditures range from 2 to 4%. Data are consistent for all three countries, resulting in roughly a quarter of costs of included diseases being attributable to overweight.

Association of Somatic Comorbidities and Comorbid Depression With Mortality in Patients With Rheumatoid Arthritis: A 14-Year Prospective Cohort Study.

OBJECTIVE: Patients with rheumatoid arthritis (RA) have a significantly increased risk of mortality compared with the general population. One of the most important predictors for mortality is somatic comorbidity. Moreover, studies have demonstrated that comorbid depression is associated with mortality. Which specific comorbidities are associated with mortality is less investigated. The purpose of this study was to investigate the association of a wide range of comorbidities with mortality in patients with RA. METHODS: Longitudinal data over a 14-year period were collected from 882 patients with RA. Data were assessed with questionnaires. The mortality status was obtained from the Statistics Netherlands for the period 1996-2011 for 99% of the patients. Somatic comorbidity was assessed in 1997, 1998, 1999, and 2008 and measured by a national population-based questionnaire including 20 chronic diseases. Comorbid depression was assessed in 1997, 1998, and 1999 and measured with the Center for Epidemiologic Studies Depression Scale. Cox regression was used to study the relationship between comorbidity and mortality. RESULTS: At baseline, 72% of the patients were women. The mean ± SD age was 59.3 ± 14.8 years, and the median (interquartile range) disease duration was 5.0 (2.0-14.0) years. A total of 345 patients died during the study period. Comorbidities that were associated with mortality were circulatory conditions (hazard ratio [HR] 1.60 [95% confidence interval (95% CI) 1.15-2.22]), respiratory conditions (HR 1.43 [95% CI 1.09-1.89]), cancer (HR 2.00 [95% CI 1.28-3.12]), and depression (HR 1.35 [95% CI 1.06-1.72]). CONCLUSION: Comorbid circulatory conditions, respiratory conditions, cancer, and depression are associated with mortality among patients with RA. Careful monitoring of these comorbidities during the course of the disease and adequate referral may improve health outcomes and chances of surviving.

Life expectancy without chronic morbidity: trends in gender and socioeconomic disparities.

OBJECTIVE: Life expectancy without chronic morbidity, or morbidity-free life expectancy (MFLE), was calculated to measure changes in population health status between 1989 and 2000 on the basis of gender and socioeconomic status. METHODS: Sullivan's method was used to calculate morbidity-free life expectancy. Prevalence rates for chronic morbidity were derived from the Netherlands Continuous Health Interview Survey. Four socioeconomic groups were distinguished on the basis of educational level. RESULTS: Between 1989 and 2000, total life expectancy increased for males and females and for all socioeconomic groups. Morbidity-free life expectancy decreased significantly for males (from 54.7 years to 53.9 years) and females (from 55.3 years to 51.0 years). The gap between males and females in MFLE has reversed, from 0.6 years in favor of females in 1989 to 2.9 years in favor of males in 2000. The gap between the upper and lower classes seems to have narrowed (for males from 11 years to 8.5 years and for females from 4.7 years to 4.0 years). CONCLUSIONS: The results indicate that morbidity-free life expectancy is falling for males and females and in all socioeconomic groups. Part of this decrease could be attributed to earlier diagnosis of chronic diseases. A widening gap in MFLE was observed between males and females in favor of males. The gap between the upper and lower socioeconomic groups seems to be narrowing.

Effect of sporting activity on absenteeism in a working population.

OBJECTIVES: To determine the effects of sporting activity on absenteeism in a working population. METHODS: Data were used from a prospective cohort study in a working population with a follow up period of 3 years and were collected with yearly questionnaires or collected from company records. Complete data on absenteeism, sporting activity, and potential confounders were collected for 1228 workers. ANOVA was used to test differences in frequency and duration of absenteeism, correlations were computed to measure the association between number of sporting years (divided by age) and frequency and duration of absenteeism, and survival analysis, according to the Cox proportional hazards model, was used to test differences in relative risk at absenteeism and recovery. All analyses were adjusted for age, gender, smoking, and alcohol consumption, and were stratified for employees with sedentary and with more active jobs. RESULTS: ANOVA showed a statistically significant higher mean duration of absenteeism among employees not practicing sports, of approximately 20 days over a period of 4 years. The survival analysis showed an increased relative risk at absenteeism (relative risk (RR) 1.09; confidence interval (CI) 1.01 to 1.18) and a decreased relative risk at recovery (RR 0.90; CI 0.85 to 0.95) for employees not practicing sports. The effect of sporting activity is larger in employees with sedentary work. No associations were found between number of sporting years and absenteeism. CONCLUSION: Employees practicing sports take sick leave significantly less often than their colleagues not practicing sports, while their periods of sick leave are shorter, especially when their work is sedentary.

Average age at first occurrence as an alternative occurrence parameter in epidemiology.

BACKGROUND: Current epidemiological methods focus mostly on incidence rates and their ratios as measures of occurrence and effect. Incidence rates and rate ratios can be hard to interpret when the outcome disease is common. METHODS: The possibilities of using an alternative measure, average age at first occurrence of the disease, are discussed. RESULTS: Methods for the analysis of average age at first occurrence are illustrated in a study of occupation and disability pension. We estimate that the average age at pension among crane drivers is only about 0.8 that of maintenance workers, regardless of the method, although the confidence intervals depend on the method. CONCLUSIONS: Average age at first occurrence is easier to interpret than the incidence rate for a very common disease. Because estimates are not readily available in software packages, we encourage further development of packaged software for this measure.

Physical activity and stroke. A meta-analysis of observational data.

BACKGROUND: Based on studies published so far, the protective effect of physical activity on stroke remains controversial. Specifically, there is a lack of insight into the sources of heterogeneity between studies. METHODS: Meta-analysis of observational studies was used to quantify the relationship between physical activity and stroke and to explore sources of heterogeneity. In total, 31 relevant publications were included. Risk estimates and study characteristics were extracted from original studies and converted to a standard format for use in a central database. RESULTS: Moderately intense physical activity compared with inactivity, showed a protective effect on total stroke for both occupational (RR = 0.64, 95% CI: 0.48-0.87) and leisure time physical activity (RR = 0.85, 95% CI: 0.78-0.93). High level occupational physical activity protected against ischaemic stroke compared with both moderate (RR = 0.77, 95% CI: 0.60-0.98) and inactive occupational levels (RR = 0.57, 95% CI: 0.43-0.77). High level compared with low level leisure time physical activity protected against total stroke (RR = 0.78, 95% CI: 0.71-0.85), haemorrhagic stroke (RR = 0.74, 95% CI: 0.57-0.96) as well as ischaemic stroke (RR = 0.79, 95% CI: 0.69-0.91). Studies conducted in Europe showed a stronger protective effect (RR = 0.47, 95% CI: 0.33-0.66) than studies conducted in the US (RR = 0.82, 95% CI: 0.75-0.90). CONCLUSIONS: Lack of physical activity is a modifiable risk factor for both total stroke and stroke subtypes. Moderately intense physical activity is sufficient to achieve risk reduction.

Avoidable mortality in Europe (1980-1997): a comparison of trends.

STUDY OBJECTIVE: To analyse international variations of trends in "avoidable" mortality (1980-1997). DESIGN: A multilevel model was used to study trends in avoidable and "non-avoidable" mortality and trends by cause of death. SETTING: Fifteen countries of the European Union, the Czech Republic, and Hungary. PARTICIPANTS: 19 avoidable causes of death among men and women aged 0-64 years. Mortality and population data were derived from the WHO mortality database; and perinatal mortality rates, from the Health for All statistical database. MAIN RESULTS: Avoidable mortality declined (1980-1997) in all the countries except Hungary. The difference between the trends in avoidable and non-avoidable mortality was small (-2.4% compared with -1.5%) and diminished over time. The largest trend variations between countries are attributable to causes mainly or partly amenable to prevention. For five of the 19 causes of death the international variations diminished over time. Various countries show trends that deviate significantly (p<0.003) from the mean trend. CONCLUSIONS: One explanation for the small and diminishing difference between avoidable and non-avoidable mortality is that some large avoidable causes show unfavourable trends. Another possible explanation is that the category of non-avoidable mortality is "polluted" by causes that have become avoidable with time. It is therefore suggested that Rutstein's lists of avoidable outcomes (1976) be updated to enable the appropriate monitoring of healthcare effectiveness. In countries that show unfavourable developments for specific avoidable causes, further research must unravel the causes of these trends.