Behavioural Risk Factors in Mid-Life Associated with Successful Ageing, Disability, Dementia and Frailty in Later Life: A Rapid Systematic Review

Background Smoking, alcohol consumption, poor diet and low levels of physical activity significantly contribute to the burden of illness in developed countries. Whilst the links between specific and multiple risk behaviours and individual chronic conditions are well documented, the impact of these behaviours in mid-life across a range of later life outcomes has yet to be comprehensively assessed. This review aimed to provide an overview of behavioural risk factors in mid-life that are associated with successful ageing and the primary prevention or delay of disability, dementia, frailty and non-communicable chronic conditions. Methods A literature search was conducted to identify cohort studies published in English since 2000 up to Dec 2014. Multivariate analyses and a minimum follow-up of five years were required for inclusion. Two reviewers screened titles, abstracts and papers independently. Studies were assessed for quality. Evidence was synthesised by mid-life behavioural risk for a range of late life outcomes. Findings This search located 10,338 individual references, of which 164 are included in this review. Follow-up data ranged from five years to 36 years. Outcomes include dementia, frailty, disability and cardiovascular disease. There is consistent evidence of beneficial associations between mid-life physical activity, healthy ageing and disease outcomes. Across all populations studied there is consistent evidence that mid-life smoking has a detrimental effect on health. Evidence specific to alcohol consumption was mixed. Limited, but supportive, evidence was available relating specifically to mid-life diet, leisure and social activities or health inequalities. Conclusions There is consistent evidence of associations between mid-life behaviours and a range of late life outcomes. The promotion of physical activity, healthy diet and smoking cessation in all mid-life populations should be encouraged for successful ageing and the prevention of disability and chronic disease.


Background
Non-communicable chronic conditions and disabilities that manifest in later life are heavily influenced by physical and social exposures, including behaviours, across the life course [1,2], leading to an accumulation of risks in older age. The four main behavioural risk factors (smoking, excessive consumption of alcohol, poor diet and low levels of physical activity) have been estimated to contribute to close to half of the burden of illness in developed countries [3,4] and are known to be unequally distributed in the population particularly affecting the most vulnerable in society [5]. Importantly there is a contingent and temporal nature of these behavioural exposures in mid-life [6][7][8] and epidemiological data suggests that it ought to be possible to prevent or delay morbidity and mortality [9] whilst promoting successful aging and quality of life [10][11][12], if adoption of healthy behaviours is encouraged across life course and society.
Although many systematic reviews have looked at the links between individual behavioural risk factors and health outcomes [13-19], a comprehensive overview across behavioural risk factors and outcomes has yet to be undertaken. There remain uncertainties regarding associations between the relative contributions of behavioural and lifestyle factors, in mid-life specifically, to prevalence, risk and outcomes of non-communicable chronic conditions (for example, dementia, cancer and cardiovascular diseases), disabilities and frailty (including quality of life, and mortality). That is particularly true for the relationship between behavioural risk factors and frailty, where the operational definition of this complex syndrome is still controversial [20,21]; and for dementia where the aetiology and natural history of disease is still uncertain [22,23].
This systematic review was one of a series of reviews conducted to inform the development of UK national public health guidance on mid-life approaches to prevent dementia, disability and frailty in later life [24]. The aim of the review was to assess the behavioural risk factors in midlife that are associated with successful ageing and the primary prevention or delay of disability, dementia, frailty (DDF) and non-communicable chronic conditions. The goal was not to summarise the whole of the epidemiological evidence on behavioural risk in adult life. Rather, we aimed to identify associations specifically derived from people in mid-life to inform the development of well-targeted interventions that will minimise the impact of ill health in later life. With a similar focus, the other two reviews in the series looked at the effectiveness of mid-life interventions on behavioural risks and late life outcomes, and at key issues for people in mid-life that prevent or limit, or which help or motivate them to take up and maintain healthy behaviours.

Methods
The review was conducted as a rapid systematic review to provide best available evidence within limited timescales. The scope of the review was defined by the funders (National of 10% of the studies was assessed twice and discrepancies resolved by discussion. No studies were excluded on the basis of quality. longitudinal cohort studies were included in the review (Fig 1). The behavioural risk factors for which we found published data in mid-life with relevant outcomes in later life were: physical activity and inactivity; diet; tobacco consumption; alcohol; weight change or weight cycling; leisure, cognitive activity or social networks; and combinations of the above. Studies of behaviour related to hearing or vision were sought, but none met the inclusion criteria. Overall, the quality of studies is good (most studies were rated as high or moderate quality). Summary quality scores can be found in Table 1. Full quality assessments can be found in S2 Table. Characteristics of included studies Summary characteristics for studies by health behaviours are listed in Table 1. Forty-five papers were found relating to mid-life physical activity (PA) or inactivity, 48 for diet, 57 for smoking and 24 for alcohol. Four studies were included that reported an association between weight change patterns in mid-life and later outcomes. Three studies reported data for combinations of lifestyle behaviours. Four studies were found that examined the relationship between midlife leisure/social activities and DDF outcomes. The data relating studies found and health behaviour is reported in Table 2. Exposures were measured a number of different ways in the included studies. For physical activity all studies used self-reports of activity except one, which used an accelerometer. Dietary data was also self-reported through interview and food questionnaires. For smoking, many studies used self-reports with questionnaires and interviews with a number of studies using biochemical testing to confirm smoking status. All studies assessing the impact of alcohol, weight change/weight cycling and leisure and social activities used self-administered lifestyle questionnaires and interviews. The outcome and exposure measures are comprehensively reported in S4 Table. Dementia & Cognition [44]; [48]; [47]; [133]; [46]; [131]; [50]; [45]; [51]; [52] [46]; [74]; [80]; [76]; [79]; [78]; [81]; [52] [26]; [126]; [134]; [125]; [129]; [52]; [132]; [127]; [124] [165]; [46]; [52]; [166]; [164]; [178] [ 184]; [46] [ 181]; [49]; [182]; [183] Overall mortality [46] No evidence was identified [46] No evidence was identified Mental health [69]; [70]; [114]; [113]; [70] [120] [ 70] No evidence was identified No evidence was identified

No evidence was identified
Only four studies explicitly examining health inequalities were identified: three looking at gender or ethnicity [26-28] and one at lower socioeconomic status groups [29].
Summary of health behaviours and outcomes. A full summary of studies, exposures and outcomes are reported in S3 Table. An overview of reported outcomes in relation to health behaviours is shown in Tables 3 and 4. Due to the large volume of data found only the main findings are reported below.

The impact of physical activity (PA) and physical inactivity (PI)
Summary data for physical activity, physical inactivity and sedentary behaviour studies, including study location, duration, period of follow-up, effect sizes, and quality scores are reported in Table 3. Note that we found no studies specifically looking at the consequences of physical inactivity or sedentary behaviour for dementia and cognition outcomes, diabetes/metabolic syndrome, or mental health. Healthy ageing and well-being. There is consistent evidence [30-32] that PA in mid-life is positively associated with healthy and successful ageing outcomes. Healthy ageing or successful survival was defined in the three included studies as having no history of major chronic diseases and no cognitive impairment, physical impairment, or mental health limitations. No studies specifically aimed to assess physical inactivity; most studies reported physical activity in quintiles so reported lower levels of PA indirectly. No studies we found assessed quality of life.
Disability/frailty. There is consistent evidence that PA in mid-life is related to more positive outcomes in terms of disability and frailty in later life. Five of the six studies reported beneficial associations between mid-life PA and physical mobility [33] or physical functioning [34] or disability [35][36][37]. One study [38] reported no significant association with disability. Another study [39] found no association between inactivity in leisure time PA in mid-life and disability at age 75.  Three studies reported on the association between mid-life PA and bone fractures or bone health. One study reported less risk of hip or wrist fractures [40,41]. One reported improved bone mineral density in those who took part in PA in mid-life [42], and another reported no significant association with the risk of osteoarthritis [43].
Dementia & cognition. There is consistent evidence that PA in mid-life is associated with lower risk of dementia or better cognitive function in later life. Of the six prospective studies reporting on dementia or Alzheimer's disease, four studies reported a significant beneficial association [44][45][46][47] and four studies non-significant associations [48][49][50][51]. Two studies found a significant inverse association between light and regular PA, but not for heavy PA [47,49]. Two studies found a positive association between mid-life PA and improved cognitive function in later life [34,52].
Overall mortality. There is consistent evidence [46,[53][54][55][56]] that regular, moderate and high intensity PA in mid-life is related to lower mortality in later life. One study [46,56] reports lower all-cause mortality but another [57] found no significant relationship between leisure time inactivity in mid-life and all-cause mortality.
Cardiovascular outcomes. Another [57] found a significant positive relationship between a single item measure of leisure time inactivity in mid-life and CVD mortality. However, no significant association was found when an index measure of leisure time physical activity was used.
Diabetes/metabolic syndrome. There is some consistent evidence that PA in mid-life is related to lower incidence of diabetes in later life [46,60,63,64]. In addition, three studies reported a beneficial association between mid-life PA and diabetes preconditions, two reported metabolic syndrome [63,65] and one reported insulin sensitivity [66].
Impact on cancer. The evidence relating to the associations between PA in mid-life and cancer is mixed. Four studies [46,53,66,67] reported longitudinal associations between mid-life physical activity and cancer or cancer mortality; however no significant relationship between mid-life PA and incident and fatal pancreatic cancer [67], lung, stomach, colorectal, lymphatic/ hematopoietic cancers [68], or cancer mortality [53,68] was observed. One study [68] found lower rates of total cancer, upper digestive tract cancers (oral, oesophagus, stomach cancer) in those who participated in moderate or vigorous PA at mid-life, and increased risk of bladder cancer in those who participated in vigorous exercise compared to those who did not. Conversely, total cancer mortality was lower in those who took part in moderate or high levels of PA [46].
Impact on mental health. The evidence for an association between mid-life PA and mental health is inconclusive. One prospective cohort study [69] reported less risk of anxiety and/ or depression for heavy PA at five-year follow-up but not at 10 years. Another study [70] found no significant association between mental well-being, including anxiety and depression and mid-life PA.

The impact of overall diet and dietary patterns
Summary data, including study location, duration, period of follow-up, effect sizes, and quality scores for overall diet and dietary patterns is reported in Table 4.
Healthy ageing/quality-of-life/well-being. There is consistent evidence [30, 71,72] that a 'healthy' diet in mid-life is related to healthy and successful ageing. A Mediterranean type diet is associated with more successful ageing [72]. A Western dietary pattern (characterised by high intakes of fried and sweet food, processed food and red meat, refined grains, and high-fat dairy products) was associated with less successful ageing. In these studies healthy ageing is defined as no major chronic diseases or major impairments in cognitive or physical function or mental health. No studies we found assessed quality of life or wellbeing.
Disability/frailty. There is limited evidence that 'healthy' diet or dietary patterns in midlife is related to better functioning. One study reported associations between mid-life diet and ADL [73], in which men who ate meat at least once every two days or more were less likely to have impairment in ADL.
Dementia & cognition. There is limited evidence that 'healthy' diet or dietary patterns in mid-life is related to dementia and cognitive functioning in later life. Two studies reported beneficial associations between fruit and vegetable intake in mid-life and dementia [46,74]. Two studies examined relationships between coffee consumption in mid-life and dementia [75][76][77]. One study reported that moderate coffee consumption was associated with lower risk of dementia, but not tea drinking [75,77]. Another found no significant association with dementia [76]. One reported a non-significant relationship between mid-life dietary antioxidant intake, flavonoids and dementia [78]. A greater risk of dementia was also reported in those consuming moderate compared to low amounts of saturated fat compared to those consuming moderate compared to low amounts of polyunsaturated fat [79].
One study reported better cognitive outcomes for those consuming a 'healthy pattern' diet [80], characterised as consumption of fruit, whole grains, vegetables, and negatively correlated with meat and poultry, refined grains, animal fat, and processed meat. More than two portions of fruit and vegetables a day was associated with better cognitive performance [52]; however, two studies reported no significant association with cognitive function [46,81]. Higher levels of total or saturated fat were associated with greater cognitive impairment in later life [82].
Overall mortality. The evidence on the impact of diet/dietary patterns on mortality is mixed. Three studies reported associations between fruit and/or vegetable intake and total mortality. One reported significantly lower risk of death in people consuming higher levels of fruit and vegetables at mid-life [83], another reported significantly lower overall mortality for each increase of 20g/day in vegetable intake [84]. Associations between >3 portions fruit and vegetables/day were not significant in one study [46]. There is limited evidence for increased mortality with greater fish consumption in those at high risk of CHD [85].
Cardiovascular outcomes. Evidence suggests a healthy diet has a positive impact on cardiovascular outcomes. Two studies reported beneficial effects of a Mediterranean diet pattern with lower risk of CHD events and mortality [86,87]. One reported fruit and vegetable intake was associated with lower CVD mortality [83]; however another [46] reported non-significant associations. Findings from one study [88] suggest that high intakes of flavonoids may be associated with decreased risk of ischaemic stroke and possibly with reduced CVD mortality.
One study [89] reported lower risk of cerebrovascular disease in those consuming meat one-two times a month compared to those consuming no meat or those who ate meat more than once a week. A lower risk of CHD events and mortality was found in women when meat was replaced with fish [90]. Another study reported no significant association between fatty fish consumption and heart failure but lower risk of heart failure in those consuming marine omega 3 fatty acids once a week [91]. Higher intakes of marine omega 3 fatty acids were not significantly associated with heart failure.
Heavier coffee drinkers showed a higher risk of CHD events and mortality compared to moderate coffee drinkers [92]. Associations were not significant for light or no coffee drinking compared to moderate intake.
A higher risk of CVD was reported for those consuming diets with the highest compared to the lowest dietary glycaemic index and glycaemic load [93]. In another study associations between glycaemic index and glycaemic load, and CVD events were not significant [91]. One study found no significant associations between mid-life protein intake and ischemic heart disease [94].
One study reported no significant associations between total, saturated, monounsaturated or polyunsaturated fat and fatal or non-fatal cardiovascular events [95].
Fruit and possibly vitamin E intake may also have protective effects against COPD [96]. In men, diets higher in fruits and vegetables and lower in meats (except fish) may reduce the risk of developing high blood pressure [97]. In women results of one study [98] suggest that higher intake of dietary magnesium may have a modest effect on the development of hypertension. Intakes of low-fat dairy products, calcium, and vitamin D were each inversely associated with risk of hypertension in middle-aged and older women [99]. While higher intake of all fruits (but not all vegetables) was significantly associated with reduced risk of hypertension in women [100]; higher intake of saturated fats, monounsaturated fats, and trans fats were each associated with increased risk of hypertension among middle-aged and older women [101].
Diabetes/metabolic syndrome. The evidence for the impact of diet on diabetes and metabolic syndrome is mixed. One study reported that a dietary pattern low in staples and high in milk was associated with lower risk of diabetes [102]. Another reported no statistically significant association between fruit and vegetables and diabetes [46]. Higher saturated fat intake at mid-life was associated with lower insulin sensitivity [66]. One study [103] reported lower risk of diabetes in women eating moderate and high amounts of fish and shellfish with a significant trend with greater fish and shellfish intake. One study found increased risk of diabetes in those consuming higher levels of red and processed meat with a significant trend from lower to higher intake [104].
Two studies reported lower risk of diabetes with coffee intake. One conducted on men and women found a significant trend towards lower risk for diabetes with increasing coffee consumption [105]. The other [106] reported a significant inverse relationship for women consuming three or more cups of coffee a day with a significant trend. In men only one-two cups/ day was significantly associated with lower risk of diabetes but there was also a significant inverse trend between coffee consumption and diabetes.
Impact on cancer. There is inconsistent evidence regarding the impact of diet/dietary patterns on cancers. One study [107] found no clear associations between four dietary patterns and cancer. Two studies reported no significant associations between fruit and vegetables and cancer incidence or mortality [46,83]. One study reported lower risk of colorectal cancer with consumption of fruit, dietary calcium, vitamin A and vitamin C; but higher risk of colorectal cancer with consumption of red and processed meat [108]. The study reported lower incidence of colorectal and rectal cancer in those consuming high volumes of milk but found no significant associations between protein and fibre intake with colorectal cancer. Another study [109] found a significant association between high salt intake and higher risk of gastric cancer in men but not women. One [110] reported lower risk of heart failure when chocolate was consumed 1-3 times month compared to no chocolate consumption. There is no significant relationship reported between flavonoids and total cancer or site-specific cancers [111].
Impact on mental health. There is limited evidence regarding the impact on mental health. One study [112] reported less psychological distress in those with the highest compared to lowest adherence to the Mediterranean diet. Light or heavy coffee consumption was also associated with lower risk of severe depression [113]. However, one [70] reported no significant association between coffee drinking and anxiety, depression or psychological symptoms, but reported lower scores on the mental health scale on the SF-36 general health questionnaire.
There was no association with tea or caffeine intake. One study [114] found no association between dietary zinc intake and depression.
Impact on other conditions. There is limited evidence regarding the impact of diet on other outcomes. One study reported intake of fruits and vegetables may reduce long-term risk of weight gain in middle-aged women [115]. Another study in women reported that weight gain was inversely associated with the intake of high-fibre, wholegrain foods but positively related to the intake of refined-grain foods [116]. Higher coffee and caffeine intake was also associated with a significantly lower incidence of Parkinson's Disease [117].

The impact of tobacco
Summary data, including study location, duration, period of follow-up, effect sizes, and quality scores for tobacco is reported in Table 3.
Healthy ageing/quality-of-life/well-being. There is consistent evidence demonstrating a detrimental association between smoking and healthy ageing, quality of life or well-being outcomes. One study [30] showed that not smoking was related to a favourable older life; another [118] found that never-smokers lived longer than heavy smokers, and their extra years were of better quality. Health-related quality of life deteriorated with an increase in daily cigarettes smoked in a dose-dependent manner. The third [119] suggests that 'ever smoking' is associated with overall survival and a borderline association with exceptional survival (i.e. free of a set of major diseases and impairments).
Disability/frailty. There is consistent evidence demonstrating the dose-response relationship between smoking and impaired mobility. One study [120] suggests that a history of smoking, especially heavy smoking, with or without quitting, is associated with an earlier onset, and more rapid development, of mobility problems during the transition from middle age to old age. Another [37] showed a consistent adverse dose-response relationship between smoking and ill-health (i.e. disability, impaired mobility, health care utilisation, self-reported health).
There is inconsistent evidence demonstrating associations between smoking and low energy fractures. One [121] showed that among women, smokers had a higher risk for vertebral fractures than non-smokers. Among men, smokers had a greater risk for low energy fractures, vertebral fractures, proximal humerus fractures, and hip fractures. Other papers showed no association between smoking and wrist fractures [40] or osteoporotic fractures [122]. One study [43,123] found a positive association between smoking and risk of osteoarthritis. No studies we found assessed frailty.
Dementia & cognition. There is strong evidence demonstrating the association between smoking in mid-life and dementia, or cognitive decline, in later life. The association between smoking and specific types of dementia is less clear. In most studies smoking was strongly associated with dementia [26, [124][125][126][127], subsequent risk of hospitalisation with dementia [26], and with being diagnosed with dementia [128]. Two studies [52,129] showed an association between smoking and cognition. One [51] showed that smoking in middle age is associated with memory deficit and decline in reasoning abilities; long-term ex-smokers (compared to current smokers and recent ex-smokers) are less likely to have cognitive deficits in memory, vocabulary, and verbal fluency. Another [130] reported greater decline in memory function, cognitive flexibility, and cognitive function among smokers. The declines in all cognitive domains were larger with increasing number of pack-years smoked. Only one study [131] found that smoking at baseline was not associated with change in cognitive decline whilst another [132] looking at dementia death failed to demonstrate an association with smoking. In another study [133], mid-life smoking was associated with an increased rate of progression of vascular brain injury, global and hippocampal atrophy.
Overall mortality. There is strong evidence demonstrating a dose-response relationship between smoking in mid-life and total mortality. Compared to never smokers, smokers are at increased risk of mortality. Five studies [118,[134][135][136][137] showed that current smokers showed the highest risks of total mortality with a dose-response relationship with increasing number of cigarettes smoked. One [135] showed that men smoking persistently were most at risk, while those who persisted in quitting had no increased risk of death compared with non-smokers. Another [138] concluded that smokers across the entire range of pulmonary function may increase their expectation of lifespan by giving up smoking. Finally, one study [139] showed that compared to current smokers, never smokers, long-term quitters and new quitters had a decreased risk of all-cause mortality; the same association was observed for never smokers and long-term quitters for other non-lung cancer mortality. Compared to those who maintained their smoking habit, individuals that reduced or quit smoking had a decreased risk of all-cause mortality [140]. Cardiovascular outcomes. Smoking or having smoked is consistently associated with increased risk of cardiovascular mortality and CVD. Six studies provide evidence of a strong association between smoking and cardiovascular mortality [137,[139][140][141][142][143], with current smokers being more likely to die from cardiovascular events compared to non-smokers. Only one study didn't support this association [144]. Other studies provide evidence of a strong association between smoking and cardiovascular events and outcomes [59,134,[145][146][147][148][149][150][151][152][153]. The highest risks for both CHD events and stroke were seen in heavy current smokers [134]. Smoking also increases the risk of CHD in men of all APOE genotypes but particularly in men carrying the e4 allele [145,154]. Only one showed no association between smoking and CHD [146]; all other studies have shown an association between smoking and stroke [59, 146,147] and myocardial infarction [148].
Diabetes/metabolic syndrome. Cigarette smoking is an independent and modifiable risk factor for type 2 diabetes; however, the evidence for insulin sensitivity and metabolic syndrome is not sufficient to conclude. Three studies [68,155,156], demonstrated cigarette smoking is an independent and modifiable risk factor for type 2 diabetes and one did not [157]. One [68] reported the risk of diabetes in those who switched from smoking cigarettes to pipe or cigars remained equal to the risk in continuing cigarette smokers. Other studies showed that smoking is a risk factor for type 2 diabetes independently of BMI and physical activity [155]; however another [157] found that smoking was associated with the metabolic syndrome but not diabetes. Finally, one study [66] found no significant association with insulin sensitivity and smoking in men.
There is some evidence to suggest that the use of smokeless tobacco in mid-life is related to type 2 diabetes [158]. The use of smokeless tobacco was associated with low insulin response but not low insulin sensitivity.
Impact on cancer. Evidence showed a consistent association between smoking and cancer with a dose-response effect. The dose-response and exposure association seems to depend on the type of cancer considered [134,159,160]; current cigarette smokers showed the highest risk of total cancer with a strong dose-response effect. In a study [161] looking specifically at pancreatic cancer in women, pancreatic cancer incidence was greater in current smokers, with the risk increasing with the number of cigarettes smoked. One [159] showed that smoking was significantly associated with colorectal cancer in men and not significantly in women. Current cigarette smoking is associated with an elevated lung cancer risk approximately 10-to 20-fold higher for squamous cell and small cell carcinoma and 2-to 3-fold higher for adenocarcinoma in both men and women [160]. Another [162] confirmed the strong association between smoking and cancer, and cancer-related mortality. Compared to current smokers, never smokers, long-term quitters and new quitters also had a decreased risk of lung cancer mortality [139].
Impact on other conditions. One study reported that being a smoker was significantly associated with weight loss [163].
No studies were identified which assessed the impact on mental health.

The impact of alcohol
Summary data, including study location, duration, period of follow-up, effect sizes, and quality scores for alcohol is reported in Table 3. Disability/frailty. Two studies reported higher odds for ill-health and osteoporotic fractures among alcohol drinkers compared to non-drinkers, while one study found no link between alcohol intake and wrist fractures [40]. Conversely, the risk for any incident osteoporotic fracture was reported to be higher among male alcohol users compared to male teetotallers [122]. A large study [37] showed that respondents with a past drinking problem had the highest odds for ill-health in terms of ADL dependence, difficulty climbing stairs, difficulty walking, poor health, and hospitalization. No studies we found assessed frailty.
Dementia & cognition. There is consistent evidence demonstrating an association between alcohol abstinence and/or heavy drinking and cognitive impairment [52,[164][165][166]. Compared to moderate alcohol intake, alcohol abstinence was associated with a higher risk of poor executive function and poor memory [52]. One study reported no association with impairment cognition or dementia [46].
Overall mortality. Evidence was mixed for associations between alcohol consumption and all-cause mortality. One study [119] reported excessive alcohol consumption was associated with overall and exceptional survival at age 85 years. One [167] showed that alcohol intake was associated with a higher risk for all-cause mortality. Another [168] showed that excessive mortality was significantly associated with heavy drinking among men diagnosed with cancer and cardiovascular disease in this time period. One reported no association between ethanol consumption and chronic obstructive pulmonary disease mortality [169].
Cardiovascular outcomes. The evidence regarding alcohol use and cardiovascular outcomes is inconsistent. Three studies [167,170,171] showed a significant association between alcohol intake and cardiovascular outcomes. Regular drinkers had a significantly lower risk of major CHD events, CHD deaths, and CVD deaths in comparison with occasional drinkers after full adjustment for lifestyle characteristics and pre-existing disease [170]. Heavy drinkers had a higher risk of both major CHD and stroke compared to occasional drinkers [167,171]. No significant associations were observed between alcohol intake and cardiovascular outcomes in three studies (e.g., disease development, death) [46,93,144].
Diabetes/metabolic syndrome. Findings were consistent with respect to the influence of alcohol intake on diabetes/metabolic syndrome. Three studies [172][173][174] found significant positive associations between alcohol intake levels and diabetes/metabolic syndrome, while one study [46] did not find an association.
Impact on cancer. There is mixed evidence demonstrating the absence of an association between alcohol intake and cancer. Three studies [46,175,176] did not find significant associations between alcohol intake and incident and fatal pancreatic cancer; cancer in general; and colorectal cancer, respectively. In contrast, another study [159] showed a significantly higher risk for colorectal cancer among men who drink alcohol compared to non-drinkers. That study [159] reported that approximately half of the reported colorectal cancer cases may be preventable by tobacco and alcohol controls in middle-aged Japanese men; however the portion attributable to alcohol is not known.
Impact on mental health. Evidence on the impact on mental health is limited. One study [70] of women indicated that past alcohol drinkers had lower anxiety than non-drinkers.
No studies were identified which assessed the impact on healthy ageing, quality of life, wellbeing.
The impact of weight change/weight cycling Summary data, including study location, duration, period of follow-up, effect sizes, and quality scores for weight change/weight cycling is reported in Table 3.
Disability/frailty. There is limited evidence that weight loss of more than 10% of max body weight in mid-life is related to hip fractures [177]. The impact was statistically significant for those aged 50-64 and 65-74 years who were weight cycling (as defined by study authors) in mid-life.
Dementia and cognition. There is limited evidence to suggest that weight change in midlife is related to dementia [178]. Those in the highest two quartiles of weight change had a significantly higher risk of dementia, independent of the direction of weight change.
Overall mortality. One study reported no association between weight cycling in mid-life and mortality [179].
Diabetes/metabolic syndrome. One study [180] reported longitudinal associations between weight change/weight cycling and diabetes. The impact was not significant for weight change or weight cycling but overall weight status was more important in those who were overweight or obese at increased risk of diabetes.
No studies were identified which assessed the impact on healthy ageing, quality of life, wellbeing, cardiovascular outcomes, cancer or mental health.
The impact of leisure, cognitive activities and social network Summary data, including study location, duration, period of follow-up, effect sizes, and quality scores from leisure, cognitive activities and social network studies is reported in Table 3.
Healthy ageing/quality-of-life/well-being. There is little evidence to determine if leisure and cognitive activities, and an individual's social network in mid-life are related to successful aging. One study [30] reported longitudinal associations between leisure, cognitive activities, social network and successful aging in both men and women. While there was a beneficial association these were non-significant. No studies we found assessed quality of life or well-being.
Dementia and cognition. There is some evidence that leisure and cognitive activities, and an individual's social network in mid-life is related to a lower risk of cognitive decline in later life. One study [49] reported an association between diversity and intensity of participation in intellectual, physical and passive activities and lower risk of AD. Three studies [181][182][183] reported longitudinal associations between leisure, cognitive activities, social network and dementia or cognitive impairment. All studies reported a beneficial association between mid-life factors and dementia or cognitive decline in later life; however in one study the associations were non-significant for some activities including social, organisational and physical activities [183]. There is limited evidence for a prospective association between mid-life mental distress and dementia [184].
Cardiovascular outcomes. We found no evidence on the association between leisure activity, cognitive activity or social networks in mid-life and late life cardiovascular outcomes apart from one study which found no correlation with reduced hypertension in women [151].
No studies were identified which assessed the impact of these behaviours on disability, frailty, overall mortality, diabetes, cancer or mental health.

Combined behavioural risks
Summary data, including study location, duration, period of follow-up, effect size, and quality scores from combined lifestyle studies are reported in Table 3.

Dementia & cognition.
There is limited evidence to suggest that combined healthy behaviours in mid-life is related to less risk of dementia in later life. Three studies [46,185,186] reported longitudinal associations between healthy behaviours and dementia or cognitive impairment; one study [185] reported that modifiable psychosocial and behavioural factors have protective effects on cognition and another [164] reported that higher levels of consumption was associated with higher risk of cognitive impairment, in particular that binge drinking was found to be an independent risk factor for cognitive impairment; however in one study the associations were non-significant [46].
Overall mortality. There is evidence to suggest that reducing unhealthy behaviours or adopting a healthier lifestyle in mid-life is related to reduced death. Two cohort studies [46,187] reported a significant negative association between number of unhealthy behaviours and mortality.
Cardiovascular outcomes. There is inconsistent evidence that reducing unhealthy behaviours or adopting healthier behaviours in mid-life is related to reduced CVD. One study reported a significant negative association between number of unhealthy behaviours and CVD [187] while the other [46] reported no association. Importantly the combined behaviours in these two studies vary.
Diabetes/metabolic syndrome. There is limited evidence that association exists between reducing unhealthy combined lifestyle behaviours in mid-life are related to diabetes. One study examined the impact of combined lifestyle on diabetes; while there was a negative correlation this was non-significant [46].
Cancer. There is no evidence that association exists between reducing unhealthy combined lifestyle behaviours and cancer. One study examined the impact of combined lifestyle on cancer and there was no correlation [46].
No studies were identified which assessed the impact on healthy ageing, quality of life, wellbeing, disability, frailty or mental health.

Disadvantaged and minority groups
For this review 'disadvantaged and minority groups' includes (but is not limited to) people of low socioeconomic status; ethnic minority groups; LGBT community groups; travellers; and other groups with protected characteristics under the equality and diversity legislation. One study examined associations between mid-life smoking and reported no difference in the risk of developing dementia by gender or ethnicity [26] or the risk of cardiovascular disease, total CHD or myocardial infarction by gender [143]. One [28] reported less risk of myocardial infarction in women participating in leisure time sports, however this result was based on a small sample size. High alcohol intake in lower SES groups is related to poorer cognitive performance [29]. No association between alcohol consumption and cognitive performance was found in those in intermediate or high socioeconomic groups. There is limited evidence and where evidence is available this is generally weak. Most of these groups also have substantially lower life expectancy than the general population, with fewer surviving into age groups of the highest risk.

Discussion
This review provides an overview of the available evidence from high quality observational studies reporting on the association between specific behavioural risk factors, in mid-life, and successful ageing and the primary prevention or delay of DDF and non-communicable chronic conditions. The conclusions of individual cohort studies are consistent with existing knowledge and best practice for a range of specific non communicable disorders (for example, reducing alcohol and tobacco consumption, and increasing physical activity and intake of fruit and vegetables).
The findings that consistently emerge as being of societal importance in their range of potential impact are physical activity, which is beneficial, and smoking is consistently found to be associated with DDF and non-communicable chronic conditions. Other factors for which the evidence base is not as strong are diet, alcohol, weight change and social/leisure activities.
For Physical Activity there was consistent evidence within the 45 studies identified that mid-life PA is associated with better late life DDF and NCC outcomes. Only two studies specifically focused on physical inactivity. This finding is in alignment with previous research that suggests regular PA can reduce the risk of chronic disease and improve one's health and wellbeing whereas a lack of physical activity leads to poorer health [188][189][190].
The review found a wealth of evidence from longitudinal cohort studies relating to the association between mid-life smoking and poorer DDF and NCC outcomes. The impact of tobacco is well known as previous research demonstrates that tobacco consumption is an important risk factor for CVD, cancer and hypertension, and is a major cause of premature mortality [191,192]. Whilst smoking cessation is associated with weight gain and a subsequent increase in risk of diabetes, the long-term benefits of giving up smoking outweigh the adverse effects of early weight gain.
There is some consistent evidence from 48 studies that healthy diets have beneficial effects on DDF and NCC outcomes and higher consumption of saturated fat or processed and red meat is associated with poorer ageing, DDF and NCC outcomes. Previous research has also demonstrated a strong relationship between diet, dietary patterns or nutrient intakes, and prevention and management of chronic diseases [193,194] including diabetes [195] and CVD [196].
Evidence from 24 studies specific to mid-life alcohol consumption was mixed and inconsistent with smaller effect sizes than for smoking and physical activity. Some studies reported negative outcomes (e.g. for dementia, mortality and cancer) and some more positive outcomes (e.g. for ageing and mental health). Previous research suggests that alcohol is a significant cause of mortality, morbidity and social problems, both in developed and in developing countries [197], causing an estimated 4% of the global disease burden including chronic diseases and certain cancers, as well as poor health in general [198][199][200].
Four studies were included that reported an association between weight change patterns in mid-life and later outcomes, including increased risk of hip fracture, mortality, diabetes and dementia. For issues specific to weight change or weight cycling it is known that being overweight increases the risks for many chronic medical conditions, including diabetes, heart disease, hypertension and stroke [201]. Modest reductions in weight can lead to important health benefits [67,202], while fatigue and unintentional weight loss can affect an individual's physical, psychological, social and cognitive functioning [203,204]. Combinations of lifestyle behaviours were also protective factors in relation to cognitive function, diabetes, vascular disease, cancer, dementia and mortality.
There is little evidence on the impact of leisure, cognitive activities and social network and combined lifestyles. Studies that focused on disadvantaged groups have tended to be few in number and of lower quality indicating that this is a neglected research area.

Limitations of the Evidence, Gaps
Limited evidence was found specifically relating to mid-life behaviours for leisure activities including cognitive activities and social networks, weight change and weight cycling and smokeless tobacco. While many diet-related studies were found they covered a broad range of diets and dietary components. There were some diets or dietary components for which studies specifically pertaining to mid-life was not found.
• Evidence of the impact of diet, consumption of dietary calcium, vitamin A, vitamin C, zinc and flavonoids on CVD mortality [83], overall mortality [84], and psychological distress [206].
No studies were found which reported: physical inactivity and healthy ageing/quality of life/ well-being, dementia, diabetes/metabolic syndrome, cancer or mental health; the relationship between alcohol consumption and healthy ageing/quality of life/well-being; weight change/ weight cycling with healthy ageing/quality of life/well-being, CVD outcomes, cancer or mental health; impacts on disability/frailty, diabetes/metabolic syndrome, cancer, overall mortality, mental health or other chronic diseases; or the impact of combined lifestyle behaviours and healthy ageing/quality of life/well-being, disability/frailty, mental health or other chronic diseases; studies of behaviour related to hearing or vision. The challenge in assessing the impact of these exposures is a pronounced lack of data.
While no included studies reported relationship between mid-life physical activity and dementia, a recent meta-analysis [212] of prospective studies was published subsequently to the completion of the review and reported a fixed effects risk ratio of 0.61 (95% CI 0.52-0.73) indicating an overall reduction in risk of Alzheimer's disease in physically active older adults.
There is also substantial heterogeneity in the measurement of exposures and outcomes included in this review making precise comparisons between studies problematic. While the review only includes longitudinal observational studies, which can show an association between mid-life risk factors and later life outcomes, associations from this type of study are not necessarily causal.
There was a paucity of data relating to disadvantaged groups covered by the equality and diversity legislation. While we found evidence assessing the risks of developing: dementia by gender or ethnicity [26]; CVD, total CHD or MI by gender [143], and in women [28] only, also alcohol intake in lower SES groups [29]. No longitudinal data was found relevant to other groups covered by the equality and diversity legislation such as LGBT groups or travellers. The absence of large-scale, high-quality research in these communities is problematic as there is a need for a more detailed and accurate picture of the disparities these communities face. Disadvantaged groups are significantly more likely to experience pervasive discrimination [213] and without clear evidence, the challenges faced are exacerbated. Whilst there have been some gains for visibility and equality for these communities, many disparities remain, particularly for women and the poorest (most vulnerable) in our society. Poorer health and well-being will persist, if high rates of discrimination are allowed to continue resulting in the experience of deeper inequality. Solutions to the disparities addressed above must focus both on reducing discrimination in general and improving conditions of living. Given that discrimination has a cumulative negative effect on health, there are plausible reasons for anticipating differences for disadvantaged groups or settings. These are the important considerations that should be made to ensure that inequities are reduced. There is therefore very little information about how midlife behaviours in disadvantaged communities' impact on population-level health outcomes in later life. More data is needed to understand how the behavioural risks are differentially modulated in these disadvantaged groups; with the aim of designing interventions that are fit for purpose.

Limitations of the Review
The remit of this rapid review was specifically to identify mid-life behavioural risk factors for DDF outcomes and common NCCs in later life. Determinants of DDF over the whole life course were not included. Pragmatically, due to the wide scope of the review, the large amount of literature covering behavioural risk factors and the outcomes of interest, and the timescales for the review, the searches were focused on studies with mid-life-related terms in the title, abstract or related MeSH indexing to identify those studies that specifically aimed to report on mid-life exposure. The implication is that cohort studies that have followed individuals from mid-to late life and reported associations of interest without specifying mid-life terms in the title or abstract were not identified by the searches. This might explain some of the gaps in evidence and further work is on-going to address this limitation.
Although we conducted a rapid systematic review it is unlikely that our overall conclusions would vary greatly had a systematic review been conducted over a longer period of time [199]. Another strength is that unlike other rapid reviews a quality assessment of the included studies was undertaken reducing the limitations associated with the evidence synthesis process and results [200].
Also, only OECD countries are included, which may impact on the implications of findings within the broader global health context. However, few studies from non-OECD countries were found in the searches.

Conclusion
The long-term impact of beneficial behavioural factors in middle to older age adults was greater chance of successful ageing and the primary prevention or delay of disability, dementia, frailty, and non-communicable chronic conditions. The exposures and health behaviours in mid-life identified in this review are important considerations both in identifying possible differential mechanisms for action (in order to reduce ill-health in the population) and the design of interventions to improve healthy behaviours.
Supporting Information S1