WWW-publications from the MORGAM Project
 |
Relative risks for stroke by age, sex and population based on follow-up of 18 European populations in the MORGAM Project: supplement to a paper published in StrokeJuha Karvanen1, Sangita Kulathinal2, Kari Kuulasmaa1, Simona Giampaoli3, Pekka Jousilahti1, Matti Niemelä1 and Kjell Asplund4, for the MORGAM Project5 |
1 Department of Chronic Disease Prevention, National Institute for
Health and Welfare, Helsinki, Finland
2 Indic Society for Education and Development (INSEED), Nashik, India
3 National Centre for Epidemiology, Surveillance and Health
Promotion,
Istituto Superiore di Sanità, Rome, Italy
4 Department of Public Health and Clinical Medicine, Medical Unit,
Umeå University
Hospital, Sweden
5 See Annex for the sites and key personnel
of contributing MORGAM Centres
|
|
© National Institute for Health and Welfare
and the MORGAM Project investigators Last updated: 18 March 2009 For more information, please contact Juha Karvanen (firstname.lastname@thl.fi) |
Document identification:
This document is a supplement to the manuscript Relative risks for stroke by age, sex and population based on follow-up of 18 European populations in the MORGAM project [1]. The supplement presents the details of the methods and the results of the analysis.
The variables used in the analysis are defined in detail in the MORGAM Manual [2]. The names of the variables and hyperlinks to their definitions are:
Variable MISSINGDATA has value 1 if at least one of the variables STROKE1, STRTIME1, BPM, BMI, DSMOKER, HDLA, and BASECVD2 has a missing value and value 0 otherwise. Percentages of risk factor measurements by population are given in Table 1.
The effect of risk factors to stroke was compared between regions, between men and women and between age-groups. The following main hypotheses were tested using likelihood ratio tests:
The testing of the hypotheses is explained in detail in Appendix 1.
On the basis of the earlier research, AGE1, BPM , BMI, DSMOKER, HDLA and NONHDL were specified as potential covariates. The regression coefficient for NONHDL was estimated to be zero and NONDHL was dropped from the model.
The validity of the proportional hazards assumption was tested using R function cox.zph [4]. Table 2 indicates that covariate BPM does not follow proportional hazards assumption. However, the correlation between between the transformed survival time and the scaled Schoenfeld residuals of BPM is only -0.05, which is quite close to zero. Therefore, all covariates, including BPM, have a linear effect to log-hazard in our models.
| rho | chisq | p | |
|---|---|---|---|
| age1 | 0.00716 | 0.161 | 0.68798 |
| bpm | -0.05363 | 9.865 | 0.00168 |
| bmi | -0.00254 | 0.022 | 0.88211 |
| dsmoker1TRUE | -0.03589 | 3.909 | 0.04804 |
| hdla | 0.01621 | 0.963 | 0.32651 |
| GLOBAL | NA | 15.333 | 0.00903 |
1 based on model: coxph(formula = Surv(strtime1, stroke1 %in% c("1", "2")) ~ age1 + bpm + bmi + dsmoker1 + hdla + strata(rua) + strata(sex), data = f31i, subset = ((basecvd2 == 2) & !missingdata))
The three main hypotheses are all hypotheses on the interactions. No additional interactions were tested or modelled.
The models are specified in tabular format where each row summarizes one model. Column "Model summary" gives a three-letter identification code with a link to the R output [3]. Column "Main effects" lists covariates that are included without interactions. Column "Interaction covariates" lists covariates that are included with interactions with the variables listed in the column "Interaction groups". Column "Strata" gives the stratifying variables in the proportional hazards model. Separate baseline hazards are allowed for categories of the stratifying variables. Column "Subset" presents the inclusion criteria for the subjects in the analysis. Column "Outcome" defines the endpoint event. The model summaries use the R syntax, which is explained in Table 3.
Table 3. R syntax for model summaries.
| Notation | Explanation |
|---|---|
| BASECVD2==2 | the value of variable BASECVD2 equals to 2. |
| & | and (logical operator) |
| | | or (logical operator) |
| ! | not (logical operator) |
| STROKE1 %IN% C('1','2') | the value of variable STROKE1 belongs to the set containing values 1 and 2. |
Models for difference between men and women
Models for difference between regions
Models for difference between age groups
Main models for fatal strokes (STROKE2)
Main models for likely cerebral infarction (STROKE3)
Comparison of models (p-values can be found here)
Figures additional to those published in StrokeFigure 1. Hazard ratios for fatal stroke during follow-up of men and women free of cardiovascular disease at baseline shown for 5 European regions and all MORGAM populations together. The horizontal lines represent 95% confidence intervals of the estimates. [png | eps]
Figure 2. Hazard ratios for ischaemic stroke during follow-up of men and women free of cardiovascular disease at baseline shown for 5 European regions and all MORGAM populations together. The horizontal lines represent 95% confidence intervals of the estimates. [png | eps]
This research was part funded through the European Community's Seventh Framework Programme (FP7/2007-2013), ENGAGE project, grant agreement HEALTH-F4-2007-201413.
