Article Text
PDF
Abstract
Objectives Our study aimed to investigate how physical activity (PA) changes over an 11-year follow-up among adults from different socioeconomic positions (SEP) near retirement age. Moreover, an analysis of different PA types is considered.
Methods We used data from the EPIC-Italy cohort. We evaluated PA using the Cambridge Physical Activity Index (CPAI) and the metabolic equivalent of tasks (MET) per hour of activity for recreational PA and household PA. Educational level was assessed using the Relative Index of Inequality (RII). Occupational classes were classified according to LIFEPATH Consortium knowledge. Logistic regression was used to analyse PA among SEP and changes during follow-up. Analyses were also conducted separately for sex.
Results The higher educated were more prevalent in the higher quartile of recreational PA than the lower educated both at baseline and follow-up (37% vs 28% and 37% vs 27%, respectively). At the baseline, the lower educated had a higher risk of being physically inactive than the higher educated based on recreational PA (overall OR: 1.50, 95% CI 1.40 to 1.60). Manual workers did not show a higher risk of less PA than professionals/managers (overall OR: 1.03, 95% CI 0.91 to 1.16).
At follow-up, the lower educated and manual workers showed a higher risk of being physically inactive (lower educated OR: 1.46, 95% CI 1.37 to 1.56; manual worker OR: 1.33, 95% CI 1.18 to 1.50). The analyses of changes in PA showed that those who were less educated or manual workers had a higher risk of worsening their PA during the follow-up period, particularly women in recreational PA and men in CPAI measurement.
Conclusion Individuals who had a disadvantaged SEP showed a higher risk of performing less PA over time.
- Physical activity
- Exercise
- Education
- Public health
Data availability statement
No data are available. Raw data cannot be made freely available because of restrictions imposed by the ethical committees, which do not allow open/public sharing of data on individuals. However, aggregated data are available for other researchers upon request to the corresponding author.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Physical activity levels differ by socioeconomic position: those in higher socioeconomic positions tend to engage in more physical activity than those in lower socioeconomic positions.
WHAT THIS STUDY ADDS
This study tries to investigate changes in physical activity over time for different socioeconomic positions.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Understanding the fluctuations in physical activity levels across various socioeconomic positions can enable the development of preventive actions toward the most vulnerable population groups.
Introduction
It is widely acknowledged that physical activity (PA) has a significant impact on both health and quality of life. Many studies have observed the benefits of PA on cardiovascular disease and mortality,1 healthy ageing,2 cognitive declines,3 diabetes,4 cancer,5 bone health6 and mental health.7 Additionally, regular PA was inversely associated with hospitalisations independently of sociodemographic and lifestyle factors.8 PA can even contribute to modulating blood pressure9 and ocular perfusion pressure.10 Overall, PA habits are an optimal indicator of the population’s well-being.
Regarding health and quality of life, differences in health according to social position have been reported consistently for several health outcomes, including self-reported health, chronic and long-term health conditions, and mortality11–13; a higher risk of cardiovascular disease14 and of several cancer types15 was found among less educated individuals.
In epidemiology, socioeconomic position (SEP) can be measured in a variety of methods. One common method is to use the level of education as a proxy when income and occupation are not available. Due to the strong influence of the family’s background and cultural resources, education also reflects childhood and adolescent SEP. Although educational attainment may be affected by illness or challenges that happened at an early age, it remains a reliable indicator of SEP.13
It has been observed SEP can have an impact on PA, as individuals with low SEP tend to be less active in adulthood.16 Furthermore, a study conducted on the European Prospective Investigation into Cancer and Nutrition (EPIC)-Norfolk cohort revealed that an increasing trajectory of PA levels over time can have a positive influence on mortality, leading to a 24% lower risk of all-cause mortality.17
However, it is still not very clear how PA changes along the life course among different social positions.18
Understanding PA trajectories during adulthood is relevant in the context of healthy ageing, and studying the determinants of changes in PA can help identify strategies to improve well-being of the most deprived elderly population.
This study aims to understand how PA changes with advancing age among different social positions, especially, when individuals reach retirement age. Moreover, an analysis of different PA types is considered.
Methods
Study population
We used data from individuals participating in the EPIC-Italy cohort. Project details are described elsewhere.19 In the period 1993–98, EPIC-Italy recruited 47 749 volunteers (15 171 men, 32 578 women, aged 35–65 years) in five different regions of the country: Varese (12 083 volunteers) and Turin (10 604) in the north; Florence (13 597) in centre, and Naples (5062) and Ragusa (6403) in the south.20 All participants in the EPIC-Italy cohort signed an informed consent form and agreed to provide information on lifestyle and nutrition and to donate a liquid nitrogen-stored blood sample. A follow-up of part of the cohort was conducted in Florence, Varese and Turin between 2004 and 2009, with an average follow-up period of 11 years.
Data collection
The lifestyle EPIC questionnaire was used to collect data20 21 that assessed sociodemographic characteristics, PA, smoking and alcohol habits, household composition, diseases, and aspects of women’s menstrual and reproductive life, following standard interviewing procedures. At baseline, the interviews were carried out in the departments of the centres involved, and questionnaires were self-administered. At follow-up, self-administered questionnaires were filled at home and returned to the centre. The diseases collected were as follows: heart attack, stroke, hypertension, diabetes, liver and kidney stones, intestinal polyps, and gastric or duodenal ulcers. During the interviews, anthropometric measurements were performed by trained interviewers using clinical equipment. At follow-up, anthropometric measurements were self-measured through a meterstick that was sent to each participant.
Socioeconomic measurement
The Relative Index of Inequality (RII) was used to measure educational level, which takes into account differences between regions, genders and birth cohorts.14 High RII values indicated a lower level of education. RII estimation can be found in Sergeant et al.22 Occupational status was harmonised into three macro-categories (according to what was done within the LIFEPATH Consortium23) by grouping classes of the European Socio-economic Classification (E-SeC).13 E-SeC is a European measure that classifies occupations into nine ordinal categories (see online supplemental material S1) based on the similarity of resources, in terms of opportunities and ‘life chances’, according to the Erikson & Goldthorpe classification.24 The three macro-categories were professionals, retailer/farmer/clerical workers, and skilled/unskilled manual workers. In addition, the category of not-employed was added to the study sample, which included subjects who answered ‘no’ to the question ‘Are you currently employed?’ on the lifestyle EPIC questionnaire at baseline.
Supplemental material
Physical activity measurement
The lifestyle EPIC questionnaire also investigated PA habits at work (occupational PA), at home (household PA) and in leisure time (recreational PA).21 25 The occupational PA was assessed using a four-category question: sedentary, standing, moderate manual work and heavy manual work, providing examples such as office worker, shop assistant, plumber and construction worker respectively.8 Recreational and household PA were assessed by questioning how many hours per week were spent in specific activities (separately in summer and winter periods) and assigning a metabolic equivalent of tasks (MET) value at each activity, using the compendium of physical activity.26
Household activities included housework, climbing stairs, do-it-yourself and gardening. Recreational activities included walking, cycling, swimming and other sports activities. Both household PA and recreational PA variables were continuous (sum of METS). Furthermore, the validated Cambridge Physical Activity Index (CPAI)25 was used to assess the overall PA level. CPAI is a four-category index (active, moderately active, moderately inactive, and inactive) calculated by combining the four categories of occupational PA (sedentary work, standing work, manual work, heavy manual work) with time spent on recreational PA (see online supplemental material S2).
Statistical analyses
The descriptive table on study population characteristics and PA were reported using frequency (and percentages) for categorical variables and mean±SD for continuous variables. Recreational and household PA variables were categorised as sex-specific quartiles. The RII was categorised as tertiles. Multivariate logistic regression was employed to analyse the relationship between PA variables (CPAI, recreational PA quartiles and household PA quartiles) and socioeconomic variables (RII and occupational status, separately) at baseline and follow-up. For logistic regression, PA variables were set as low vs high. The low CPAI category grouped ‘inactive’ and ‘moderately inactive’ of CPAI. Low recreational PA category and low household PA category grouped the first and second quartiles for recreational PA and household PA, respectively. Analyses were conducted on the overall cohort and separately for men and women. The logistic model was adjusted for age at recruitment, sex (except for sex-stratified estimates), EPIC centre, presence of disease (yes/no), Body Mass Index (BMI), and alcohol and smoking habits. The adjustment set was chosen consistently with other studies on the EPIC cohort.8 9 Subjects with missing values on PA variables and socioeconomic variables at baseline or follow-up were removed. ORs and their 95% CI were reported.
To assess whether individuals improved or worsened their PA during follow-up, a variable was created by making the difference between PA at follow-up and PA at baseline (CPAI and recreational PA, respectively). Values ranging from −3 (maximum worsening) to +3 (maximum improvement) were obtained for each observation. A value of 0 (zero) indicated that the individual maintained their PA level over time. However, only those with the lowest measurements at baseline could obtain the maximum improvement values (+3). Conversely, only those with the highest levels of (PA) at baseline could obtain the maximum worsening values (−3). Consequently, a value of −3 (maximum worsening) is no worse than a value of −1 (minimum worsening). For this reason, we dichotomised the difference variable into worsening (values = −1, −2, −3) vs maintenance or improvement (values=0, 1, 2, 3) to apply the logistic model.
Possible interactions between variables in the models were explored: variables that showed high interaction were treated separately. The chi-square test was performed. The test was two-sided and was considered a p value <0.05 to be significant.
BMI was categorised into three categories: up to 24.9 (normal weight), 25–29.9 (overweight) and 30 or more (obese).27 28
Additional sex-stratified analysis on workers aged 50 or over was conducted to assess the relationship between PA and SEP in individuals entering retirement age during the follow-up. In the 2000s, the retirement age in Italy was 65 for men and 60 for women.
Sankey diagrams were built using sankeymatic.com to show how subjects moved between different categories of CPAI from baseline to follow-up. Analyses were performed with STATA18.0.
Equity, diversity, and inclusion statement
The group of authors comprises 12 members, including seven women and five men. They consist of research fellows, junior, mid-career and senior researchers from various disciplines such as physiotherapy, epidemiology, hygiene, and biostatistics. However, all members of the author group belong to the same country.
Our research study involved males and females from different socioeconomic backgrounds, ages, and clinical and demographic characteristics. The (PA) outcomes were determined by considering various settings. We have discussed the influence of gender in our study. However, we acknowledge the lack of generalisability of our results due to some cohort characteristics, which we have explained in the discussion.
Results
After removing missing values (n=4004, 14.4%), this study included 23 734 participants with data both at baseline and follow-up. Baseline characteristics are presented in table 1. With an overall mean age of about 50.5 at baseline, women made up 68% of the sample. Specific PA characteristics at baseline and at follow-up are reported in table 2 and in table 3. At CPAI, the lower educated showed a higher percentage of being active than the higher educated. Consequently, skilled/unskilled manual workers were the more prevalent occupational group in the active category. Regarding recreational PA, the higher educated were more prevalent in percentage in the higher quartile than the lower educated, both at baseline and follow-up (37% vs 28%, 34% vs 25%, respectively). Sex-specific PA characteristics at both baseline and follow-up were reported in online supplemental material S3 and S4. There was a slight deterioration in the CPAI from baseline to follow-up, as shown by using the Sankey diagram (figure 1). Study population movements on CPAI between baseline and follow-up for each tertile of RII were reported in the online supplemental materials S5–S7.
Baseline characteristics of study population
Baseline physical activity frequencies of study population by groups
Follow-up physical activity frequencies of study population by groups
Sankey diagram, Cambridge Physical Activity Index flows from baseline to follow-up (FU).
Table 4 presents the findings of the logistic regression on the CPAI. At baseline, overall ORadj showed that lower-educated were less likely to be physically inactive than higher-educated (OR 0.71, 95% CI 0.66 to 0.76). Regarding the occupational classes, the skilled/unskilled manual workers class showed a 51% reduced probability of being physically inactive compared with the professionals/managers class (OR 0.49, 95% CI 0.43 to 0.55), while the clerical/retailer/farmer class had a higher risk of being physically inactive compared with the professionals/managers class (1.23, 95% CI 1.10 to 1.38). Sex-stratified estimates followed the pattern of the overall estimate: the skilled/unskilled manual workers class showed a lower risk of performing less PA than the professionals/managers both for men (OR 0.40, 95% CI 0.33 to 0.47) and women (OR 0.62, 95% CI 0.52 to 0.74).
Association between socioeconomic characteristics at baseline and low (vs high) PA level as assessed by Cambridge Physical Activity Index (CPAI) at baseline and follow-up (low CPAI level = inactive, moderately inactive; high CPAI level (ref.) = active, moderately active).
At follow-up, results on RII were in the opposite direction than at baseline: the lower-educated were more likely to be physically inactive than the higher educated (OR 1.20; 95% CI 1.11 to 1.30). The sex-stratified model showed similar results to the overall ORsadj for women but not for men. Regarding occupation, the class of clerical/farmer/retailer workers and the not-employed had a higher risk of being inactive than professionals/managers. Sex-stratified results for women showed clerical/farmer/retailer, skilled/unskilled, and not-employed had a higher risk of being physically inactive than professionals/managers (OR 1.48, 95% CI 1.24 to 1.77; OR 1.31, 95% CI 1.08 to 1.58; OR 1.83, 95% CI 1.52 to 2.21; respectively).
Table 5 shows the results of PA changes according to CPAI. There, the lower-educated had a higher risk of worsening over time than the higher-educated (OR 1.49, 95% CI 1.39 to 1.60). In sex-stratified estimates, lower-educated men had a 92% higher risk (1.92, 95% CI 1.70 to 2.17) of worsening in CPAI than higher-educated. Women showed a milder risk increase (OR 1.32, 95% CI 1.22 to 1.44). Regarding occupational classes, the skilled/unskilled manual workers class showed a higher risk of worsening PA than the professionals/managers class (OR 1.92, 95% CI 1.69 to 2.18), especially for men (OR 2.20, 95% CI 1.85 to 2.61).
Association between baseline socioeconomic characteristics and changes in Cambridge Physical Activity Index (CPAI) over time (follow-up minus baseline) (worsening in CPAI level (−3 to −1) vs maintenance/improvement in CPAI level (0 to +3))
The results of recreational PA at baseline and follow-up are reported in table 6. Subjects who were lower-educated had a higher risk of being physically inactive (based on recreational PA) than the higher educated, both at baseline (1.50, 95% CI 1.40 to 1.60) and at follow-up (1.46, 95% CI 1.37 to 1.56). Lower-educated women showed a greater increase in risk than men compared with higher-educated educated. For occupational status, it was important to note how skilled/unskilled workers and not-employed subjects had a higher risk of being physically inactive than the professionals/managers of follow-up (1.33, 95% CI 1.18 to 1.50; 1.36, 95% CI 1.19 to 1.54, respectively). In the sex-stratified analysis, every other woman’s occupational category showed a higher risk of being physically inactive in recreational PA than the professionals/managers class.
Association between socioeconomic characteristics at baseline and low (vs high) PA level as assessed by recreational PA at baseline and follow-up (low recreational PA=I and II quartiles; high recreational PA (ref.) = III and IV quartiles)
In table 7, the ORadj of changes in recreational PA over time showed an increased risk of worsening recreational PA for the lower-educated compared with the higher-educated (OR 1.09, 95% CI 1.01 to 1.16). In sex-stratified analyses for RII, less-educated men showed a higher risk of worsening recreational PA than more-educated men. Regarding estimates for occupational status in the overall cohort, skilled/unskilled manual workers and not-employed showed a higher risk of worsening their recreational PA than professionals/managers (OR 1.29, 95% CI 1.13 to 1.47; OR 1.47, 95% CI 1.28 to 1.69; respectively). In sex-stratified analyses, manual workers and not-employed women had a higher risk of worsening their recreational PA than professionals/managers (OR 1.41, 95% CI 1.16 to 1.70; OR 1.57, 95% CI 1.30 to 1.90; respectively).
Association between baseline socioeconomic characteristics and changes in recreational PA (follow-up minus baseline) (worsening in recreational PA level (−3 to −1) vs maintenance/improvement in recreational PA level (0 to +3))
Results on household PA are reported in online supplemental material S8. Overall, baseline and follow-up estimates showed that lower-educated were less likely to be physically inactive at home than higher-educated, for both men and women (baseline: OR men 0.61, 95% CI 0.54 to 0.69; OR women 0.45, 95% CI 0.42 to 0.49. Follow-up: OR men 0.64, 95% CI 0.56 to 0.72; OR women 0.64, 95% CI 0.59 to 0.69). Regarding occupational status, professionals/managers seemed to be the least active category in household PA. The further analysis conducted on workers aged 50 or more (online supplemental materials S9–S12) showed similar results to the analyses presented above: the lower-educated and manual workers had a higher risk of being physically inactive than the higher-educated and the professionals/managers (respectively), both at baseline and at follow-up. In addition, manual workers show a higher risk of worsening overall PA (based on CPAI) and recreational PA than the professionals/managers.
Discussion
In this large prospective study conducted on the EPIC-Italy Cohort, we found that individuals with higher levels of education or those from less disadvantaged social classes had higher levels of PA than others and better maintained the level over time, which is especially visible when work activity ceases, particularly, in women. Our findings are consistent with prior studies regarding the relationship between PA and SEP and between PA and occupational status.29–38
Although at baseline the less educated or those belonging to lower social classes engage in more overall PA than the more educated individuals, the less educated tend to perform low PA over time, especially, when reaching retirement age. This was further supported by the analysis conducted on the difference between follow-up and baseline, where it was observed that people with a higher inequality index had a higher risk of worsening their PA levels in the follow-up period. Regarding occupational classes for both men and women, those in the manual worker class were shown to have a significant risk of worsening PA levels over time compared with the professionals/managers class.
Based on the CPAI, individuals with lower levels of education presumably had higher baseline levels of PA since they were more likely to do PA during work hours but less likely to do so during leisure time.33 However, the reasons for this occurrence may also be attributed to other social factors. According to Jaana T. Kari and colleagues,31 education is related to decision-making abilities, which can lead individuals to make healthier long-term decisions in their behaviour. Additionally, there may be a mediating effect of income on the relationship between education and PA: a higher education increases income levels, which, in turn, provides greater opportunities to invest in PA. In accordance, findings of our cohort showed the less educated were less likely to do recreational PA at both baseline and follow-up, and a higher risk of worsening recreational PA levels over time than the more educated.
The WHO defines PA as ‘any bodily movement produced by skeletal muscles that requires energy expenditure, […] all movement including during leisure time, for transport to get to and from places, or as part of a person’s work’.39 However, the quality of PA can play a significant role in health outcomes. Recent studies observed that workers who perform more PA at work have worse health outcomes than those who perform less physically demanding jobs40 and that higher levels of PA at work can lead to an increased risk of cardiovascular disease, in contrast to leisure-time PA.41 This occurrence has been described in the literature by Holtermann42 as ‘the physical activity paradox’. According to the author, a possible explanation for the discordance of different PA settings on health outcomes could lie in the substantial difference between the two types of activity (recreational PA vs occupational PA). Occupational PA is usually characterised by a sustained low intensity over a long period with inadequate breaks and recovery times, whereas, recreational PA is usually carried out over a shorter period with a moderate to vigorous intensity and adequate resting time. Additionally, levels of occupational PA appear to be inversely correlated with recreational PA among individuals: those who showed higher levels of recreational PA had sedentary jobs, while those who had heavy manual jobs showed lower levels of recreational PA.43 44
Our results show that skilled/unskilled manual workers tend to do less recreational PA than managers/professionals at follow-up but not at baseline.
Contrary to our expectations, the results of the sex-stratified models for recreational PA on men were not statistically significant compared with women. However, several studies have observed that men tend to engage in more PA during life than women independently of education and equity.32 33 45 In addition, we had a higher proportion of women than men in our sample. This could account for our results (see table 1). However, the lower-educated men tended to worsen the recreational PA level over time in our cohort.
Another aspect worth considering is the effect of diseases on PA habits. Diseases may hinder PA among subjects, in a pattern of reverse causality. Dong Hoon Lee and colleagues27 in their study emphasise how recent disease events or undiagnosed illnesses may overestimate the effect of PA on health and mortality. However, this effect occurred when the follow-up of studies was shorter. In our study, the follow-up is over 10 years on average.
The difference that emerged between the clerical/farmer/retailer class and the skilled/unskilled manual worker class deserves further investigation. The class of clerical, farmer and retailer workers showed a lower PA level at baseline than the professionals/managers class (based on CPAI, which also considered PA at work). One possible explanation could come from the fact that, in comparison to the farmers and retail workers, the second occupational class consists of 89% of clerical workers who perform sedentary work (see table 1).
In this study, household PA was also analysed. In the literature, studies concerning the effect of household setting are not yet very consistent. However, similar to recreational PA, it has been observed that higher levels of household activity may be associated with a reduction in the risk of all-cause mortality and mortality from cardiovascular and respiratory causes, irrespective of leisure-time PA.46 47 In addition, higher levels of household activity seem to be associated with better subjective well-being, and a higher volume of grey matter in the elderly.48 49
However, the habit of household PA among different levels of socioeconomic position is still unclear.48 As mentioned, higher SEP was associated with higher overall levels of PA.16 48 However, in our study, the relationship between household PA and SEP is the opposite of the one described in the previous sentence (Supplementals): less-educated individuals tend to perform more household PA than more educated individuals, particularly, in the female sex. The higher prevalence of women in PA habits at home has also been found in other studies.50 51 Further investigation of the effects of household PA on health outcomes would be necessary. However, this study may offer new insights into the habit of household PA among SEP.
One of our study’s strong points is its prospective design. Second, the analysis included a large number of subjects (n=23 734) with a long-duration follow-up which, given the mean age of the population, allowed us to study PA levels in people entering retirement age. However, there are a few limitations. First, although the lifestyle EPIC questionnaire has been validated for a different country,25 52 PA was measured on self-reported data. Second, the cohort examined, while very large and varied, is not representative of the general Italian population. This is because the subjects in the cohort are volunteers with good lifestyles and on average, more aware regarding eating habits than the general population at the time. However, it is not a very physically active cohort. The type of data available represented another limitation of the study: whereas, recreational PA and household PA were quantitative ordinal variables per observation (MET), occupational PA, measured with Cambridge Physical Activity Index, was a four-categorical variable. Moreover, CPAI did not represent a pure assessment of occupational PA but a combination of activity at work (categorical) and recreational PA. The use of other validated questionnaires to measure moderate-to-vigorous physical activity in adults would have allowed a more in-depth evaluation.53 54 However, most of these evaluation surveys were validated after the start of the data collection process of the EPIC cohort. Finally, due to observational design, residual confounding cannot fully be discharged. Despite the abovementioned point, our study represents one of the largest studies that assessed changes in PA through various SEP measurements.
Further studies should explore the mediating role played by diseases on PA changes, and how PA changes may predict the onset of new diseases.
Conclusion
The study’s findings indicate that individuals from lower social positions had an increased risk of being physically inactive over time than individuals from higher social positions, particularly, as they reach retirement age, mainly concerning recreational PA. Those with higher levels of education or social position were more likely to engage in recreational PA, which represents the only form of PA that is protective against all causes of mortality. These results can be considered a subject for public health discussion to develop preventive actions toward the most disadvantaged individuals in the community.
Data availability statement
No data are available. Raw data cannot be made freely available because of restrictions imposed by the ethical committees, which do not allow open/public sharing of data on individuals. However, aggregated data are available for other researchers upon request to the corresponding author.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants. The EPIC study has been approved by the Ethical Review Board of the International Agency for Research on Cancer (IARC, Lyon) and by the local Ethical Committee. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
The authors wish to acknowledge and thank Franco Merletti for his useful comments on the manuscript. The first author’s work was part of her research project for the master’s degree in Epidemiology at the University of Turin.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
SC and FR contributed equally.
Contributors MF conceptualised and designed the project, managed and analysed data, and drafted the paper. FR and SC contributed to conceptualisation and design, provided the data, monitored the statistical analyses, and revised and refined the paper, with both contributing equally. LF contributed to monitoring the statistical analyses and the management of the data. GM contributed to the conceptualisation and design, statistical analyses, provided the data, and read and revised the paper. CS, ML, LD, BB, MA, VV and PV read, reviewed and refined the paper. MF is guarantor. IA was not used at any stage of the project.
Funding EPIC-Italy was funded by the Italian Association for Research on Cancer (AIRC, Milan). The funding agency played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.