Cardiovascular disease (CVD) is the leading cause of death worldwide. According to the Global Burden of Disease study, which compiles and analyzes population health data from 204 countries, the number of CVD cases almost doubled between 1990 and 2019, rising from 271 million to 523 million worldwide.
At the same time, CVD-related mortality has risen steadily from 12.1 million deaths to 18.6 million over the same period.
In order to prevent and treat these diseases, it is essential to take gender into account. CVD in women is often misdiagnosed, underestimated and undertreated. The incidence of CVD also varies according to sex and age: it is higher in adult men, while it increases in women after menopause.
However, recent research suggests that gender also has an effect on the incidence of CVD, hence the need to consider it in carrying out research.
Our team brings together researchers from several disciplines, including occupational health and safety, epidemiology and population health.
Specifically, we focus on the individual and organizational factors that influence workers’ health. Public health agencies like the World Health Organization, governments, funding agencies and scientific journals are increasingly demanding that research in public health and other fields take sex and gender into account in order to better understand health inequalities between women and men.
Inequalities in research
It has been known for several years that the incidence of CVD varies according to sex and age. It is higher in adult men, for example, but increases in women after menopause.
Inequalities have also been observed in CVD research, which mainly focuses on men. As a result, women are underrepresented, a situation that can compromise the effectiveness of the prevention strategies in place. This highlights the need to examine gender differences in the development of CVD.
However, in addition to sex, gender could also influence the effectiveness of existing prevention strategies, according to various studies.
Sex and gender
Sex and gender, although different, are closely related and both influence health.
Sex, which refers to a set of biological attributes, can put women at a higher risk of stroke at different times in their lives, for example during pregnancy and menopause. Gender, on the other hand, is neither innate nor static. It is a dynamic concept, linked to social roles, expectations and behaviours.
The results of a recently published study by our research team on the association between gender and the incidence of CVD suggest that gender influences CVD risk in different ways, depending on sex.
In fact, we found that men with certain characteristics traditionally associated with women (for example a lower level of education, having a job characterized by few physical demands) have a 42 per cent increase in the risk of CVD compared to men with characteristics traditionally associated with men. This result highlights the importance of prevention strategies that take both sex and gender into account.
However, it’s difficult to consider gender when using existing data, because the data rarely include a direct measure of gender. To address this shortcoming, we propose a new measure of gender to be used in health studies that use secondary data.
Gender-related variables
Researchers have in the past suggested ways to combine gender-related variables. However, our team has found that these measures have significant limitations, which we are trying to overcome in our study. For example, one of these measures includes variables that are not usually measured in health studies (such as the level of responsibility for care and the level of discipline of the children in the household).
This makes it more difficult for this measure to be generalized and reused by other researchers. Another measure has included variables that, like gender, are likely to be factors of discrimination, such as sexual orientation or citizenship. But it’s believed that including these variables could lead to an erroneous classification of individuals, and therefore distort the measurement and its associations with health.
To address these issues, we have developed a gender index that allows this dimension to be integrated later into databases that do not take it into account, while circumventing the limitations mentioned above. Our study also aimed to see if this index is associated with the risk of cardiovascular disease, regardless of gender.
Our research team proposes a four-step method to create this index.
1. Selection of gender-related variables
To identify gender-related variables in our database, we used a four-dimensional definition, as suggested by the Canadian Institutes of Health Research.
Here are the dimensions and self-reported variables that we considered:
1) Gender roles: Marital status, hours worked per week, family responsibilities, physical and psychological demands of the job and decision-making latitude at work
2) Gender identity: Personality traits (anger, cynicism and hostility)
3) Gender relations: Social support outside and at work
4) Institutionalized gender (Which refers to the way power and resources are distributed according to gender): Level of education and job category
Our initial position was to assume that these variables are related to gender, because they are traditionally different for women and men.
2. Confirmation of a final list of gender-related variables
To confirm the relevance of the variables included in our list, we have kept only those that predict sex.
The reason is simple: gender-related characteristics are influenced by social norms and expectations. These are generally attributed to men and women in very different ways and change over time, as well as from one culture to another.
Therefore, by applying the LASSO method in a logistic regression model we were able to, 1) exclude the factors that are less relevant, and 2) increase the accuracy of our prediction.
At the end of this stage, we excluded the hostility variable. By removing the hostility variable, our model became more reliable in differentiating between the sexes. This variable either blurred the results or did not really help. Without it, the model became more accurate and coherent.
3. Assignment of a gender score
We then calculated a gender score to classify the participants. This calculation was made according to the male and female characteristics confirmed in the previous step.
It indicates the probability of being a woman (between 0 and 1) for each respondent. Lower scores are interpreted as more masculine traits, while higher scores reflect a higher level of feminine traits.
In other words, this gender measurement is seen as a continuum from masculinity (gender scores around 0) to femininity (gender scores around 1). Subjects with a score close to the average represent the androgynous gender type in this respect.
This measurement is divided into three groups (or terciles), according to predetermined thresholds. Individuals are thus grouped in tercile according to their predominantly masculine (T1), intermediate (T2) and feminine (T3) characteristics.
4. Validation of the gender measurement
Finally, to validate our measurement, we examined the correspondence between gender and sex.
As expected, the majority of men were in tercile 1, and most women in tercile 3. However, some men and women did not fit this pattern, which shows that sex and gender are not completely linked. They are therefore partially independent, as some people had gender scores that did not fully correspond to their sex.
The use of this index by different researchers could allow them to highlight gender-related differences that could not have been observed based solely on sex.
Moving towards greater equality in research?
Our measure fills some gaps in the consideration of gender in health studies using secondary data.
Including both sex and gender in health research could make prevention strategies more suitable for everyone. To achieve this, studies that reuse existing data could draw on our method to create gender measurements adapted to their culture and context.