Approaches to studying behavioural clusters

Health behaviours have been defined as overt behavioural patterns, actions and habits that relate to health maintenance, to health restoration and to health improvement^{(Reference Gochman28)}. In the literature, these are usually considered in a broad or narrow context. Behaviours sensu largo include a whole spectrum, including diet, physical activity, substance use, sexual activity, environmental hazard avoidance, preventive health and safety practices, e.g. seatbelt use^{(Reference Harris and Guten29)}. Health behaviours sensu stricto are usually limited to behaviours often abbreviated as SNAP: smoking, nutrition, alcohol and physical activity⁽³⁰⁾ also known as the big four⁽³⁾.

Two methodological approaches in studying the clustering of health behaviours can be considered: a priori and a posteriori. The first approach is hypothesis-driven. Based on the previous knowledge, the predefined dietary and lifestyle risk factors are selected and its cumulative effect on health is investigated. The results of studies which applied this approach were the keystone in public health research, documenting the detrimental effects of the main four modifiable lifestyle behaviours on health⁽³¹⁾. As can be predicted, the cumulative effect of all four behaviours (SNAP) generates the highest health risk^{(Reference Khaw, Wareham and Bingham32)}. However, it has not been clearly specified which components of these behaviours in particular would create the clusters (patterns) of the highest or lowest risk to health.

To answer this research question, the a posterori approach needs to be applied. This means that the patterns of behaviours are derived with the use of multidimensional statistical techniques based on the available data, rather than the researcher's hypothesis^{(Reference Armitage, Berry and Matthews33,Reference Wirfält, Drake and Wallström34)} . Most commonly applied techniques include factor analysis, latent class or principal component analysis. Next, the associations are being investigated between the derived clusters (commonly named patterns, e.g. dietary patterns or dietary-lifestyle patterns) and health outcomes using modelling, e.g. regression analysis or survival analysis^{(Reference Wadolowska, Hamulka and Kowalkowska35)}. The exploratory nature of this technique means that initially the components (sets of variables) are not known to the researcher and no assumptions can bias the results. This approach helps to obtain objective, often unexpected results which could not be predicted and investigated with the use of an a priori approach. This approach received a great deal of interest in the past decade, documenting the specific dietary patterns in various subpopulations^{(Reference Kant36,37)} . However, the evidence regarding the clusters of combined dietary and lifestyle behaviours remains scarce, particularly in young men. In this study, we will focus on the published research, which applied the a posteriori approach, often referred to as the empirical method.

Clustering of dietary-lifestyle behaviours in young men

A recent review pooled data from fifty-six studies in adults (both men and women)^{(Reference Noble, Paul and Turon21)}. The authors found that being a male, with lower socioeconomic status and younger age is associated with adherence to riskier clusters. Often, those risky clusters had a common denominator: alcohol use and smoking. The authors also highlighted that diet was less often included in the reviewed studies in comparison to alcohol use and smoking. Moreover, the dietary component was often limited to one or two variables, e.g. the frequency of fruit and vegetable consumption or breakfast skipping. Interestingly, the associations with physical activity were unclear. In general, poor diet tends to cluster with substance use (alcohol, tobacco and/or drug use). Clusters differed depending on the sample (e.g. general population v. students), country and analytical methods used^{(Reference Noble, Paul and Turon21)}.

The reasons behind the clustering can be broken down into three segments: cognitive, social and biological^{(Reference Patterson, Haines and Popkin38)}. The cognitive reasons are linked to the theory of reasoned action and determinants such as intentions, self-efficacy, outcome expectancies, perceived susceptibility and perceived severity^{(Reference Conner, Norman, Conner and Norman39)}. It has been previously documented that men display a lower level of nutritional knowledge than women^{(Reference Whitcombe, Nethery and Bergman40)}. Furthermore, in men, nutritional knowledge is not a direct predictor of healthy eating attitudes^{(Reference Brown, McIlveen and Strugnell41,Reference O'Brien and Davies42)} . This would explain why some behaviours in men remain inconsistent, for example, as mentioned earlier, there is no clear association between physical activity and diet or substance use; men can have unhealthy dietary habits and at the same time be highly active^{(Reference Noble, Paul and Turon21)}. Gubbels et al.^{(Reference Gubbels, Kremers and Stafleu43)} identified a cluster in boys labelled as Sports-Computer, which was characterised by health-promoting dietary behaviours, sedentary time and using a computer as well as being physically active. Although the authors studied behaviours in a sample of children, this example shows that some behaviours that may appear to be contradictory often coexist, and this phenomenon can be revealed using the data-driven approach.

Social reasons for the clustering of behaviours are related to age, sex, socioeconomic status, education, and these determinants appear to be the most frequently studied in lifestyle research^{(Reference Lv, Liu and Ren19,Reference Meader, King and Moe-Byrne44,Reference Murphy, MacDonncha and Murphy45)} . These will be discussed in more detail further in the review. Lastly, biological explanations are associated with physical and mental health. It has been suggested that sex-related differences in lifestyle behaviours might be linked to hormonal differences^{(Reference Courtenay46)}. Higher levels of androgens and cortisol, and lower serotonin levels in men may increase the desire for intense sensations and experiences, which can be obtained with the use of stimulants^{(Reference Courtenay46)}. It has been shown that men have different responses to stress than women, with one of the avoidant coping strategies is alcohol or substance use^{(Reference Courtenay46)}. The psychological aspects of lifestyle behaviours have been previously described by Conry et al.^{(Reference Conry, Morgan and Curry47)} Among six identified clusters, authors found that the multiple unhealthy behaviours reflected in ‘multiple risk factor’ and ‘physically inactive’ clusters, more prevalent in menaged 18–29 years, were associated with lower levels of energy and vitality and highest psychological distress^{(Reference Conry, Morgan and Curry47)}.

Due to the cross-sectional design of these studies, the results need to be cautiously interpreted. The potential reason can be either a cause or a result of certain behaviours, e.g. psychological distress can be a cause of unhealthy lifestyle choices but also a result of long-term poor lifestyle habits.

Clusters of dietary-lifestyle behaviours and cardiometabolic health

Although a number of studies have investigated the clusters of lifestyle behaviours in mixed-sex population, the number of studies in young men which used empirical methods and included a dietary component is very small. A study by Lawrence et al.^{(Reference Lawrence, Mollborn and Hummer48)} investigated the associations between lifestyle clusters and cardiovascular health in a large American cohort. Eight indicators were used to create a cardiovascular risk scale (systolic and diastolic blood pressure, pulse rate, Hb A1c, TAG, HDL- and LDL-cholesterol, BMI). The results revealed that two clusters named ‘inactive, poor diet, poor sleep’ and ‘consistently negative’ were the predictors of poorer cardiovascular health in young adults (aged 26–31) when compared to the ‘consistently positive’ cluster. However, these results were not sex-specific and only two dietary variables were used to assess the diet quality: frequency consumption of fast-foods and sugary beverages^{(Reference Lawrence, Mollborn and Hummer48)}.

Another interesting study, by Greene et al.^{(Reference Greene, Schembre and White49)}, investigated the clustering of diet, physical activity, psychosocial factors and elevated health risk associated with obesity. The authors found that male students (aged 18–24) from the cluster labelled as ‘psychosocially secure’ had better adiposity outcomes than men from other clusters, despite the lowest intake of fruit and vegetables and the lowest level of physical activity^{(Reference Greene, Schembre and White49)}. This paradox again shows that investigating only one aspect of a diet (here mean daily fruit and vegetable intake) may be insufficient to reflect overall diet quality.

In our recent study in young men (age <40 years), we proposed to include a broader spectrum of dietary variables along with lifestyle behaviours, such as physical activity at work and free time, current and past smoking, alcohol use and screen time^{(Reference Lonnie, Wadolowska and Kowalkowska50)}. Using a data-driven approach, we identified four dietary-lifestyle patterns. The early cardiometabolic risks associated with adiposity were observed in two dietary-lifestyle patterns labelled ‘sandwiches and convenient diet’ and ‘fast-foods and stimulants’, while two other dietary-lifestyle patterns appeared to be protective; due to its components, these were labelled ‘protein food, fried-food and recreational physical activity’ and ‘healthy diet, active, past smokers’. Both of these dietary-lifestyle patterns were associated with lower odds of obesity and lower risks of elevated total cholesterol (only ‘protein food, fried-food and recreational physical activity’) and fasting blood glucose (only ‘healthy diet, active, past smokers’), despite some unhealthy components such as frequent consumption of fried foods or smoking in the past. It could be that the physical activity component present in both patterns, alleviated the effect of the unhealthy behaviours^{(Reference Lonnie, Wadolowska and Kowalkowska50)}. This study, however, highlights that studying the dietary-lifestyle patterns using the empirical method mirrors the behaviours truly existing in populations (rather than compliance to guidelines recommendations). We believe that this approach more closely represents the real-life scenario, i.e. studies have shown that only a small percentage of the population entirely follow dietary recommendations, whereas a higher percentage display more relaxed, yet still pro-healthy dietary behaviours that can still be protective and used as an example for the rest of the population. There is an increasing recognition among academics and practitioners that health guidelines and interventions should be informed by the evidence obtained for the behaviour patterns research⁽⁵¹⁾. Researching patterns, rather than individual health-related behaviours, recognise the fact that lifestyle behaviours are complex and may have synergistic or antagonistic effects on the human body.

Practical implications

Modifying lifestyle behaviours in this population pose a serious challenge. Men are described as a population not only hard-to-reach but also hard-to-engage^{(Reference Lefkowich, Richardson and Robertson52)}. It is also unclear whether single or multiple health behaviour change interventions are more effective^{(Reference Prochaska and Prochaska53)}. In the scoping review, Ashton et al.^{(Reference Ashton, Hutchesson and Rollo54)} identified forty-six interventional studies which have targeted risk behaviours solely in young men (age 17–35 years) and fifty-four studies reporting outcomes with stratification for age and/or sex. Only a fraction of studies targeted poor diet (n 5), with the majority of the interventions being focused on increasing physical inactivity, lowering alcohol use and reducing risky sexual behaviours^{(Reference Ashton, Hutchesson and Rollo54)}. The important limitation in terms of applicability of the results was low diversity. Most often, the samples consisted of university students (being a population easily accessible to the researchers). Based on the studies included in the review, the authors concluded that the effectiveness of future interventions can be improved by: (i) being tailored to the specific subpopulations of men (e.g. unemployed, ethnic minorities, etc.); (ii) accounting for the different types of masculinity; (ii) incorporating theoretical frameworks; (iii) choosing the right delivery channel (face-to-face or online intervention), as well as intervention facilitator, similar to the targeted group^{(Reference Ashton, Hutchesson and Rollo54)}. Football Fans in Training (https://ffit.org.uk/) is a good example of a successful, age- and sex-sensitised, health-promoting initiative ^{(Reference Hunt, Wyke and Gray55–Reference Bunn, Wyke and Gray57)}. The 12-week intervention directed to overweight men aged 35–65 was delivered in Scottish football clubs by community coaches and involved dietary guidance and physical activity sessions^{(Reference Hunt, Wyke and Gray55)}. Participating men managed to implement lasting lifestyle changes, lose bodyweight, improve blood pressure and increase self-esteem^{(Reference Hunt, Wyke and Gray55–Reference Bunn, Wyke and Gray57)}. The appeal and success of the intervention were attributed to feeling similar to other men in the group, sharing similar interests, filling sessions with effervescent experiences and receiving continuous peer-support^{(Reference Bunn, Wyke and Gray57)}.

Getting to know the target population is fundamental. The European Commission on Men's Health has stressed the fact that behavioural sex-related differences represent different needs and perceived barriers in terms of health promotion^{(Reference White, De Sousa and De Visser58)}. In further work, Ashton et al.^{(Reference Ashton, Hutchesson and Rollo59)} suggested that focus groups might be helpful in identifying young men's motivators and perceived barriers towards eating healthily and being active. In this group of 18–25 years-old men, the main motivators for healthy eating were physical health, sport or performance, physical appearance and social influences, while key motivators for physical activity were physical appearance, social inclusion, physical and mental health and improvements for sport or performance^{(Reference Ashton, Hutchesson and Rollo59)}. Hence, it may be concluded that the ‘one size fits all’ approach is not feasible and a thorough analysis of the targeted population is recommended to design the intervention accordingly to the specific needs.

Recommendations for future studies

In the majority of studies, the dietary component was often limited to one or two variables, most commonly the frequency of fruit and vegetable consumption. It can be argued that the intake of fruit and vegetables is a reasonable indicator of diet quality. It is the food group included in all dietary indices designed to assess diet quality^{(Reference Burggraf, Teuber and Brosig60,Reference Waijers, Feskens and Ocké61)} and its high intake is associated with the reduced risk of CVD, cancer and mortality^{(Reference Wang, Ouyang and Liu62)}. Hence, it might be presumed that high fruit and vegetable intake is a key marker of healthy dietary pattern. This has been previously confirmed by Ramsay et al. ^{(Reference Ramsay, Shriver and Taylor63)} in a study in children. Those who consumed whole fruit, fruit juice and non-starchy vegetables (apart from white potatoes) had higher diet quality measured with healthy eating index (HEI-2010)^{(Reference Ramsay, Shriver and Taylor63)}. However, the results from this sample cannot be generalised to other populations and older age groups. Moreover, studying only a singular dietary behaviour reflects the level of adherence to that particular dietary recommendation and does not expose the nuances in the complexity of human behaviour. As shown before, young men may engage in a mixture of explicit dietary behaviours^{(Reference Greene, Schembre and White49)}_, and therefore in this case, the examination of a singular dietary habit does not represent the overall diet quality. Lastly, from the public health perspective, examining solely fruit and vegetable intake or breakfast skipping will not highlight other areas which require improvement and should be addressed in the target population.

Similarly, variables such as smoking, alcohol use or physical activity would benefit from a more in-depth assessment if used as variables in data-driven analyses. A detailed description of those behaviours could add to the existing knowledge, by revealing what specific types of lifestyle behaviours generate the highest risks and which ones might be protective, e.g. smoking (past, current, frequency, tobacco or electronic cigarettes), alcohol consumption (type, frequency, amount per drinking occasion), physical activity (recreational v. physical labour, moderate v. vigorous intensity, etc.).