Interpretation of the results
The results were in line with the first hypothesis, which assumed that educational inequalities would be larger and more distinct for preventable cancers. Indeed, the majority of
preventable cancer sites showed higher mortality rates for low-educated men. In general, cancer sites amenable to
behavioural change more often showed significant educational inequalities compared to cancer sites amenable to
medical interventions. In Belgium, health insurance is mandatory, and covers about 99% of the total population [
39]. This explains why inequalities are larger for the cancer sites related to behavioural change instead of to medical interventions.
Cancers of the lung and the head and neck showed the largest educational inequalities both in absolute and relative terms. Generally, these results are consistent with other European studies. Mackenbach et al. [
16] also observed the largest relative educational inequalities in Europe in the 2000s for cancers of the head and neck, lung and oesophagus [
16]. In France, the largest inequalities were observed for the same cancer sites, except for colorectal cancer characterized by modest relative inequalities [
1]. Inequalities were also pronounced for cancers of the oesophagus and liver. This was in line with a study of Jemal et al. [
40] covering inequalities in the 1990s and 2000s in the United States. Other European studies found similar results, with inequalities in cancers of the lung, upper aero-digestive tract, and stomach explaining a great deal of the SEP inequalities in general cancer mortality [
2,
3,
41].
The second hypothesis, which assumed an increase in inequalities for preventable cancers, was not confirmed by our results. Absolute cancer mortality generally showed a trend towards less inequality. In addition, the cancers that did show an increase in mortality differences, i.e. cancer of the colorectum, pancreas, central nervous system and malignant melanoma, were only partly preventable (colorectal cancer and malignant melanoma). Stomach cancer mortality showed the largest decrease in inequality, both in relative and absolute terms. The decrease in stomach cancer mortality was larger among the low-educated groups, which points to the fact that the advancement that was made (the decline in prevalence of Helicobacter pylori infection [
42] was now widespread in society [
14]). Studies examining recent trends in educational inequalities for multiple cancer sites using nationwide exhaustive population data are scant to our knowledge. Nationwide studies based on sample data in France and Britain [
1,
43] and on linked mortality data in Barcelona [
41] also showed generally stable relative inequalities in male cancer mortality over time. However, in France, absolute inequalities declined for men [
1], as observed in our study.
Link with fundamental cause theory: Differences in resources
Educational differences in mortality reflect differences in resources [
16]. Education implies knowledge resources that can be utilized to maximize health [
44]. These resources include a variety of capacities such as financial means [
45]; stable employment [
44,
45]; health literacy [
45]; being receptive to prevention messages [
46]; being able to change health behaviours [
46]; and making proper use of the health system [
46]. Consequently, it does not come as a surprise that the cancer sites with the largest educational differences are those that are highly amenable to behavioural change (e.g. cancers of the lung or head and neck) and (to a smaller extent) cancer sites amenable to medical interventions (e.g. colorectal and prostate cancer), which is in line with the fundamental cause theory [
16,
23,
24]. As educational inequalities are observed for almost all cancer sites, we can assume that there is not one single cause in terms of proximal factors that can be responsible for these inequalities [
44]. Consequently, both disease risk factors and factors related to healthcare should be taken into account [
44].
Low educated people are more vulnerable to unhealthy behaviours such as smoking, physical inactivity, being overweight and obese, (excessive) alcohol consumption, bad oral hygiene, risky sexual behaviour, human papillomavirus (HPV) infection and exposure to occupational agents [
16,
45,
47,
48]. Interactions between these risk factors even strengthen their carcinogenic effects [
41]. Low educated people are also more likely to be in bad health initially, and prevention messages about healthy habits and collective facilities (e.g. tobacco control initiatives) might have a differential impact among the social strata [
44,
49]. Moreover, low educated people are more likely to have lower levels of social support and to have less control over their lives [
4]. Another important risk factor for (cancer) mortality is healthcare utilization. This is especially important for cancers amenable to medical interventions, as these have a 5-year relative survival rate of more than 70% [
27]. Low educated people are less likely to seek timely medical attention (causing late stage at diagnosis), and are less likely to have access to good quality healthcare [
3,
16,
24,
43,
45]. Likewise, participation rates in organized screening are lower among low educated people [
50]. Moreover, high educated people are more likely to be early adopters whenever new developments in disease management are made [
19,
44,
50]. Taken together, low educated people might be more susceptible to new arising health threats, and hence show higher cancer incidence rates, as well as lower survival rates due to a lower ability to cope with the aggressiveness of cancer and respond to the treatment [
3,
19]. Despite the almost full coverage of health insurance in Belgium, differences in health care utilization are still observed by SE group. Data of the Belgian Health Interview Survey (BHIS) prove that low educated men are more likely to smoke, to have limited physical activity, to be obese and to show excessive consumption of alcohol [
51‐
54]. No educational gradient has been observed in the participation rate of colorectal screening [
55], however, low educated Belgian men were more likely to delay medical care because of financial reasons [
56].
The cancers showing (large) educational inequalities in this study are all associated with lifestyle-related factors. We will now discuss the most important ones.
Lung cancer is widely acknowledged as being caused by smoking tobacco [
1‐
3,
5,
6,
29,
41,
50] as well as occupational exposures [
3,
57,
58]. The observation that lung cancer mortality declines in all educational groups points to the fact that Belgian men went through all four phases of the smoking epidemic [
20]. Head and neck cancers (oral cavity and lip, larynx and pharynx) are associated with smoking, as well as with alcohol use [
1,
2,
48,
50,
59‐
61]. Earlier research reported that these lifestyle habits are causing 70% of head and neck cancers [
48,
61], with alcohol being the most important contributor [
50]. Another recently emerging risk factor for head and neck cancers (especially oropharyngeal cancer) is infection with HPV [
61‐
64]. Yet, according to the literature, a considerable part of the burden and aetiology of cancers of the head and neck remains unexplained [
48,
61]. Colorectal cancer is associated with behavioural factors such as cigarette smoking, alcohol use, physical inactivity, and excess bodyweight [
41,
65,
66]. Mortality decreased in all educational groups, probably due to a healthier lifestyle [
49], and to better treatment protocols [
49]. Targeted screening, hereby reducing late stage at diagnosis, might also contribute to the decline [
65]. Despite the overall decrease in mortality, the decline takes place at a faster pace for high educated men, resulting in increasing absolute educational inequalities. Although the BHIS did not observe educational differences in colorectal cancer screening [
55], educational differences in stage at diagnosis (because of postponement of seeking medical help due to financial reasons) cannot be ignored as a possible explanation as well as differences in lifestyle factors [
51‐
53]. Stomach cancer mortality has significantly dropped over time, yet educational differences remain. Stomach cancer is an aggressive tumour with a short survival. This suggests that the inequalities are mainly due to exposure to risk factors, rather than to differences in healthcare [
41,
67]. Risk factors associated with stomach cancer are infection with Helicobacter pylori as well as smoking [
41,
67]. Educational inequalities in mortality due to cancer of the bladder and oesophagus remained important. Again, smoking (for both), and alcohol use and dietary disorders (for oesophageal cancer) are likely to play a part [
1,
41,
50]. Although prostate cancer mortality declined in all groups, low educated men are still worse off relative to high educated men. The most important risk factor is high age [
67]. Since prostate cancer has a high survival rate, educational differences are probably related to differences in management. Advancements have been made in disease management (e.g. hormone therapy, a wider adaptation of radical prostatectomy in the elderly, prostate-specific antigen test, and radiotherapy) [
67], and high educated men might be more likely to use these developments.
For the
non-preventable cancers, inequality is rising for cancer of the pancreas and central nervous system. Although not defined as preventable, apart from old age, genetic factors and medical conditions (such as diabetes mellitus, chronic pancreatitis, or cholecystectomy), the only established risk factors are cigarette smoking (explaining only one fourth), and food consumption patterns [
68,
69]. A large part of the aetiology thus remains poorly understood [
69]. Pancreatic cancer is a relatively rare tumour but because of its extremely low survival rate, mortality rates are quite high [
69]. Based on the assumption that innovation in the prevention and treatment of pancreatic cancer is lacking [
6], we would assume the association between education and pancreatic cancer mortality not to have changed over time. However, a reversal of the association was observed, with a disproportional decrease in pancreatic cancer mortality in favour of high educated men. This trend might be related to the smoking epidemic.
Strengths and limitations
This study evaluates (trends over time in) educational inequalities in cancer mortality. The dataset used for this study consists of a high-quality, exhaustive dataset containing all deaths during the study period. To our knowledge, such a rich source of information containing nationwide individually linked data on cancer mortality and educational attainment is unique outside the Nordic context. Moreover, through the direct individual link between census and register data, a numerator-denominator bias was eliminated. This high-quality standard of the dataset enables to give precise estimates of (trends in) the association between cancer and education at the individual level. In order to capture the full extent of inequalities and to avoid bias, both absolute and relative measures of inequality were calculated [
31‐
34].
A limitation of this study is that the dataset only provides information on mortality, and not on cancer incidence and survival. Cancer mortality being the result of cancer incidence and cancer survival, this paper only tells part of the story [
41]. Furthermore, the dataset does not contain any information on health behaviours or access to and quality of healthcare. The reasons behind the cancer mortality inequalities thus largely remain a black box (cf. infra).
SEP was operationalized using educational attainment. Related to health, education captures a person’s capacity to prevent health damage and to tackle illness through suitable care pathways [
3]. Education is commonly used as an indicator for SEP and has many advantages [
19]. It is available for almost everyone in the population [
44‐
46], in contrast to job status, which is more difficult to capture for the retired and non-working population [
44]. Furthermore, education is a stable indicator and has a close association with other indicators of SEP [
45], such as job status and income [
19]. Moreover, compared with other SEP indicators, education is less sensitive to reverse causation, as it is obtained relatively early in life [
16]. A disadvantage is that it is related to both age and period [
46]. Our data showed an increase in the share of the population that is highly educated. To account for the different educational distribution between both periods, RIIs were estimated [
5]. However, we must bear in mind that the different educational distribution might reflect a shift in the role and significance of education which we cannot adjust for [
44].
Furthermore, a transition took place in the ICD coding system between the two periods: from ICD-9 to ICD-10. This can possibly account for some of the variation in mortality rates between the periods, although its impact is probably limited since differences between the revisions are minor [
2,
6,
17,
49].