Edited by: Roberto Cattivelli, Italian Auxological Institute (IRCCS), Italy
Reviewed by: Marta Bastos, Universidade São Judas Tadeu, Brazil; Ena Monserrat Romero Pérez, University of Sonora, Mexico
This article was submitted to Health Psychology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The aim of the study is to determine the association between Behavioral Lifestyles (regular physical activity, healthy diet, sleeping, and weight control) and longevity in the elderly. A search strategy was conducted in the PsycInfo, Medline, PubMed, Web of Science (WoS), and Scopus databases. The primary outcome was mortality/survival. Four variables (mean of participant's age at the baseline of the study, follow-up years of the study, gender, and year of publication) were analyzed to evaluate the role of potential moderators. Ninety-three articles, totaling more than 2,800,000 people, were included in the meta-analysis. We found that the lifestyles analyzed predict greater survival. Specifically, doing regular physical activity, engaging in leisure activities, sleeping 7–8 h a day, and staying outside the BMI ranges considered as underweight or obesity are habits that each separately has a greater probability associated with survival after a period of several years.
In general, it is genetic and environmental factors which are held to be the main determinants of survival and longevity (Christensen and Vaupel,
In this sense, the research project “
A brief historical portrait combining
Perhaps the clearest perception of the changes happening across longevity is provided in the survival curves by Roser (
According to Riley (
This extraordinary demographic change: mortality reduction, increasing survival, longevity, as well as life expectancy at birth and throughout the life cycle, occurred in historical association with socio-economic growth, advances in bio-medical and social development, and health and social care, as well as higher education and the general improvement in living conditions.
The WHO (
Taking into account an aging of aging (aging plus?) panorama, the WHO not only presents useful demographic projections for planning the near future but speculates about how
The WHO thus posited a new view of
In sum, as already stated, the main hypothesis of the
There is no academic definition of
It is worth remembering when healthy lifestyles started. Socio-historical and demographic events began to lead to an extraordinary growth of life expectancy in the middle of the nineteenth century. At this time, mortality in women was very high because of death from puerperal fever. Dr. Philipp Semmelweis, a Hungarian gynecologist, was working in a Viennese hospital. He observed that gynecologists went directly from the dissection room to the delivery room without taking any antiseptic measures. He then discovered that the incidence of puerperal fever could be drastically cut by the use of hand disinfection in obstetrical clinics when the obstetric doctor and matron started washing their hands carefully before proceeding to the birth.
A century after those events, the concept of lifestyles or healthy life habits gained widespread importance through the link to the Alameda County Study, which started in 1965 and has continued in several waves until the present day. It was designed to investigate normal daily habits, including social relationships, in order to detect risk factors for poor health and mortality in everyday life. The first wave comprised 6,928 participants who completed questionnaires and were followed at intervals for up to 20 years after the initial investigation.
The study yielded
The
Some criticism has also been voiced, as has been emphasized by many authors, Alameda County lifestyles are strongly linked to the social contexts of the USA (see Hankin,
The objective of this meta-analysis is to synthesize the results of primary studies that evaluate, through longitudinal designs, the relationship between mortality/survival and four of the main lifestyles considered healthy. These four styles have in common that they are observable through objective measures and that they are modifiable: (1) Regular physical activity; (2) Weight control; (3) Healthy diet; and (4) Sleeping 7–8 h per night. To do this, combined estimates of the effect size are obtained, and their significance is evaluated. In addition, the potential moderating role of four variables related to the study design and the characteristics of the participants is analyzed.
This report follows the guidelines of the APA task force recommendations about reporting standards for quantitative research in Psychology, and especially in meta-analysis articles (Appelbaum et al.,
A systematic search was performed on five websites that provide access to multiple databases related to academic articles. The websites reviewed were PsycInfo, Medline, PubMed, Web of Science (WoS), and Scopus, up to April 2021. The searching strategy was guided by a specific question: Which healthy lifestyle factors are related to longevity in the elderly? Four healthy lifestyle factors were considered as behavior lifestyles: (1) Regular physical or cognitive exercises; (2) healthy diet; (3) sleeping 7 h; (4) weight control. The selected keywords were: “longevity,” “life expectancy,” “lifestyle,” “healthy lifestyle,” “physical activity,” “exercise,” “diet, healthy,” “body weight,” “sleep hygiene/classification,” “sleep/epidemiology,” “sedentary behavior,” “longitudinal studies,” “follow-up studies,” “prospective studies,” “twin studies,” “meta-analysis,” and “elderly” or “aged.” No time restrictions were imposed. The words selected were introduced as free terms and they were searched in the title, abstract, and keywords boxes. The searching strategy is available in
The inclusion criteria were as follows: (1) the study assessed the association between the selected healthy lifestyle factors on longevity; (2) the target population of the study was focused on the elderly; (3) the study involved outcome indicators as a measure of one or more of the four healthy lifestyle factors; (4) the design of the study was a prospective study; (5) the study reported statistical results on the association between healthy lifestyle factors and mortality or survival. The exclusion criteria were as follows: (1) the study did not address specifically mortality or survival; (2) the study did not address any of our selected healthy lifestyle factors; (3) the design of the study was not longitudinal; (4) the mean age of the sample population was not elderly, over the study period. Additionally, if two studies were based on the same dataset, even partially, the study with more follow-up years was selected. When a study did not report the statistic that reflects the association assessed and their confidence interval, an exact
Flow diagram of the study selection.
Database searches were conducted in April 2021. Potentially eligible studies were selected in two steps; the first step was based on the title and abstract screening. Irrelevant references were removed. The second step was based on the full-text reading of potentially relevant studies. The pre-specified eligibility criteria were checked in both phases. For each reference, the following variables were systematically extracted, and they were entered into a summary table: (1) Author, year; (2) number of follow-up years; (3) sample size at the baseline; (4) percentage of female participants; (5) mean age at baseline; (6) overall death rate; (7) death rate of the reference group; (8) Dataset or project name; (9) participants (10) predictor variables; (11) outcome; (12) Assessment/Instrument(s); (13) Index type [Hazard ratio, odds ratio (OR), relative risk (RR)]; (14) Effect size; (15) Confidence interval of the effect size. The collected data is available through the author's mail account.
Regular physical exercises, healthy diet, sleeping 7 h, and weight control were the healthy lifestyle factors selected due to their significant association, according to previous literature, with a higher probability or a larger longevity. Also, those factors have been widely studied and are supported by several previous studies.
There were several measures reported by the selected studies. Specially, several ways of operationalizing and quantifying the variables of interest were observed. It should be noted that, the categories of the studies were based on those measures.
Altogether, the physical activity has been measured in various ways, by the practice/absence of the physical activity in the daily routine of the participant, by the frequency of the physical activity practiced (times per week), by the intensity of the activity performed (metabolic equivalent of task—METs), by taking walks on a regular basis, or engaging leisure activities. Based on that criterion, we have done a sub-division within the regular physical exercise factor (see
Means values (and range) in the four moderators studied, grouped according to the factors assessed.
PA (Yes-No) | 71.8 |
52.4 |
13.9 |
2006 |
|
PA frequency | 65.5 |
56.8 |
14.9 |
2008 |
|
Vigorous PA | 63.1 |
51.6 |
13.1 |
2013 |
|
Leisure activity/leisure physical activity | 64.9 |
46.2 |
13.4 |
2014 |
|
Walking | 69.7 |
52.5 |
9.1 |
2006 |
|
59.7 |
51.3 |
13.6 |
2012 |
||
76.3 |
54.2 |
10.8 |
2009 |
||
Obesity | 64.6 |
48 |
13.8 |
2009 |
|
Overweight | 60.8 |
42.5 |
14.1 |
2009 |
|
Underweight | 64.5 |
54 |
12.8 |
2008 |
In the case of the Healthy Diet factor, some studies used the adherence of participants to the Mediterranean diet as a measure, other studies reported the frequency or number of daily servings of fruits and vegetables (WHO recommendations), and other studies focus on measuring the amount of calories consumption. In all studies, the control group was determined by the lowest adherence to the Mediterranean Diet (score), the lowest frequency of fruit and vegetable intake or the lowest number of servings; instead of, the focus group was determined as the opposite category to the control group, for example the highest frequency of fruit and vegetable intake.
Respect of the Sleeping 7–8 h factor, two of the three selected studies established 7–8 sleeping h as the control group and one of them used 6 h as control group; and a greater number or lesser number of sleeping hours, than control group, was considered the focus group.
Regarding the Weight control factor, with body mass index (BMI) as a proxy, the categories used refer mostly to the limits established by the World Health Organization, although in other studies the authors considered the division into quantiles or cut-off points based on the median or mean of the complete sample or by gender sample. For our purpose, a sub-division of Weight control was assumed as follows:
It must be mentioned that the selected studies reported analyses based on one or more of the four factors of healthy lifestyle studied in this meta-analysis. In the studies that reported separate estimates by gender, they were taken as two different samples in the same study. Sixty-one studies assessed regular physical activity, 14 studies focused on healthy diet, 3 studies assessed sleeping 7–8 h, and 36 studies assessed control weight (see
The assessments done in the studies and their instruments of measure are described in
Lifestyle indices scores were created in some primary studies. Those indices usually combined two or more of the factors that are examined here. As their combination rarely match, it is not appropriate to synthesize these combined indices. We have only worked with the indices referred to the factors separately for those studies.
The effect size index in this meta-analysis is the
Variability between studies was evaluated using the
Meta-regression moderator analyzes were performed to assess four potential sources of heterogeneity: participants' mean age at baseline, length of follow-up, gender (percentage of women), and year of publication.
The risk of publication bias, as reflected in the asymmetry of a funnel plot, was assessed by visual inspection of the figure and some statistical tests, as the Egger's test, the rank correlation test, and the Trim and Fill method. We have also calculated the fail-safe numbers (Rosenthal,
Searching databases resulted in 655 unique records. Of these 655 articles, 357 were excluded in the first step, based on the title and the abstract (
The median year of publication was 2012 (range 1990–2020). The median sample size was 6,382, ranging from 148 to 654,827. The median of the average age was 64.5 years (38.7–93.1 years). Studies were conducted in Europe (36.6%), United States (34.4%), Asia (12.9%), United Kingdom (9.7%), Australia (2.1%), Central America (1.1%), South America (1%), and United States and Europe at the same time (2.1%). The median follow-up time was 12 years, varying between 2 and 46 years (see
The means and ranges in the four moderators of the included studies are shown in
Association between physical activity, frequency of physical activity, and vigorous physical activity, and the risk of mortality.
The forest plots with the studies involved in Walking and Leisure Activities/Leisure Physical Activities are shown in
Association between walking regularly, and performing leisure time activities, and the risk of mortality.
Regarding the studies referred to the “leisure activities/leisure physical activity,” the control group was integrated by people who rarely perform this type of activity. The results showed a significant association between frequent leisure time activities and/or leisure physical activities and mortality [HR = 0.81; 95%CI: 0.72–0.90], that means participants who frequently engage in leisure time activities and/or leisure physical activities get a mortality risk of 19% lower than people who rarely or never perform those activities. Again, it should be noted that the heterogeneity in both factors is high [QWalking (12) = 29.471,
The forest plot with the studies involved in the healthy diet factor are presented in
Association between having a healthy diet and the risk of mortality.
Association between hours of sleep and risk of mortality.
Body mass index was considered as a proxy for weight control. The forest plots with the three groups of BMI are shown in
Association between obesity, overweight, and the risk of mortality.
Four variables (mean of participant's age at the baseline of the study, follow-up years of the study, gender, and year of publication) were analyzed to evaluate the role of potential moderators. Obviously, we have not analyzed the potential role of moderators in those factors whose effect is not significant. That is, it was analyzed for all factors except overweight.
Analysis of publication bias associated with significant estimates.
PA (Yes–No) | 2,127 | 90 | 0.690 | 0.549 | 0 | HR = 0.75 |
PA frequency | 916 | 85 | 0.675 | 0.361 | 2 | HR = 0.76 |
Vigorous PA | 8,923 | 205 | 0.099 | 3 | HR = 0.69 |
|
Walking | 810 | 100 | 4 | HR = 0.76 |
||
Leisure activity/leisure physical activity | 325 | 75 | 0.153 | 1 | HR = 0.82 |
|
Healthy diet | 2,058 | 115 | 0.455 | 0.793 | 0 | HR = 0.85 |
Sleeping | 70 | 65 | 1.000 | 4 | HR = 0.90 |
|
Obesity | 1,028 | 190 | 0.618 | 4 | HR = 1.19 |
|
Underweight | 1,897 | 145 | 0.931 | 0.774 | 1 | HR = 1.41 |
In addition, the studies involved in the “PA Vigorous” dimension reported measures based on the PA intensity in some cases, and measures referred to the PA quantity in other cases. When the intensity of the physical activity was informed, it usually was measured in METs (PA is estimated as the energy cost of a given activity divided by resting energy expenditure), and the PA quantity was linked to the number of hours per week, hours per month or kilocalories consumed for the subject during a regular physical activity. For this reason, two categories were considered, one called
On the other hand, the association between the mortality rate and frequently engaging in leisure time activities, including or not physical activities, is moderated by the average age of participants at baseline. In other terms, performing leisure time activities has less relation to the mortality rate (the association is lower) the older the person is at the beginning of the study.
For the Sleeping factor, the length of the follow-up interval and the year of publication of the study are moderators of the association between the mortality rate and sleep 7 or 8 h. In the case of follow-up years, the negative slope indicates that the association is smaller the more years it is followed during the study. The slope is positive for the year of publication and indicates that the association is greater in the most recent studies.
Regarding the BMI, the only significant moderator of the association between the mortality rate and overweight is the length of the follow-up years of the study, which shows a positive slope, which means that the association is greater the greater the follow-up years. It should be noted that the association between overweight and the rate of mortality was not significant.
The results of the moderators should be interpreted cautiously, since the high number of tests (40) predicts that, assuming a significance threshold of α = 0.05, in the absence of any effect 5% of the tests will be significant (approximately 2, in this set of tests). In this case there are more significant results (6; 15%), but some of them are probably mere type I errors.
The well-known tendency to facilitate the publication of significant results over non-significant ones can produce an over-estimation of the effect size, or even the appearance of an artificial, non-existent effect. We have evaluated the degree to which this anomaly, known as publication bias, could be a potential threat to the results of this meta-analysis. In this respect, we have made visual inspection of the funnel-plot figures and we have tested for asymmetry with the Egger's test, the rank correlation test, and the Trim and Fill method. We have also calculated the fail-safe numbers for those effects that are significant. The results are summarized in
Results of meta-regression models on lifestyle factors.
PA (Yes–No) | MAB | 16 | −0.0033 | −0.014 to 0.0074 |
FUY | 15 | 0.0071 | −0.0041 to 0.0182 | |
GENDER | 16 | −0.0012 | −0.0049 to 0.0025 | |
PUBY | 16 | 0.0024 | −0.0126 to 0.0175 | |
PA frequency | MAB | 13 | −0.0029 | −0.0163 to 0.0106 |
FUY | 13 | −0.0082 | −0.0276 to 0.0113 | |
GENDER | 12 | 0.0000 | −0.0056 to 0.0055 | |
PUBY | 13 | −0.0053 | −0.0267 to 0.0162 | |
Vigorous PA | MAB | 34 | 0.0023 | −0.0038 to 0.0084 |
FUY | 33 | 0.0013 | −0.0058 to 0.0084 | |
GENDER | 35 | −0.0007 | −0.0017 to 0.0017 | |
PUBY | 36 | 0.0018 | −0.0081 to 0.0117 | |
Walking | MAB | 13 | −0.0031 | −0.0115 to 0.0052 |
FUY | 10 | 0.0417+ | −0.0046 to 0.088 | |
GENDER | 13 | −0.003 to −0.0004 | ||
PUBY | 13 | −0.0038 | −0.0149 to 0.0072 | |
Leisure activity/leisure physical | MAB | 11 | −0.0202 to −0.0001 | |
activity | FUY | 11 | 0.0154+ | −0.0011 to 0.0319 |
GENDER | 11 | −0.0032 | −0.0085 to 0.0028 | |
PUBY | 12 | −0.0038 | −0.0257 to 0.0116 | |
Healthy diet | MAB | 21 | −0.0013 | −0.0055 to 0.0029 |
FUY | 21 | −0.003 | −0.008 to 0.0021 | |
GENDER | 21 | −0.0004 | −0.0013 to 0.0016 | |
PUBY | 21 | 0.0074+ | −0.0013 to 0.0161 | |
Sleeping | MAB | 7 | 0.0044 | −0.0024 to 0.0111 |
FUY | 7 | −0.2887 to −0.0427 | ||
GENDER | 7 | −0.0000 | −0.0077 to 0.0068 | |
PUBY | 7 | 0.0057 to 0.0428 | ||
Obesity | MAB | 29 | −0.0094+ | −0.0198 to 0.0010 |
FUY | 32 | 0.0150 | −0.0031 to 0.0330 | |
GENDER | 32 | 0.0006 | −0.0019 to 0.0030 | |
PUBY | 32 | −0.0136 + | −0.0294 to 0.0022 | |
Overweight | MAB | 29 | −0.0094+ | −0.0198 to 0.001 |
FUY | 28 | 0.0009 to 0.0193 | ||
GENDER | 29 | −0.0018 + | −0.0036 to 0.000 | |
PUBY | 29 | −0.0047 | −0.0168 to 0.0075 | |
Underweight | MAB | 23 | −0.0051 | −0.0117 to 0.0014 |
FUY | 25 | −0.0071 | −0.0266 to 0.0123 | |
GENDER | 26 | −0.0011 | −0.003 to 0.0009 | |
PUBY | 26 | 0.0007 | −0.0114 to 0.0128 |
The table shows that all the results are safe from this threat, in the sense that, although the effect may have been over-estimated, at least it exists and is not a by-product of selective publication, as all fail-safe numbers exceed Rosenthal's criterion. Asymmetry tests reveal no evidence of publication bias. The Trim and Fill test has imputed several values to some of the funnel plot. However, in no case has the estimate ceased to be significant after the imputation of values.
The results of the present meta-analysis allow us to reliably conclude that lifestyles are associated with mortality (survival). The four factors studied, which are usually identified as healthy lifestyles, allow us to predict greater survival. Specifically, doing regular physical activity, engaging in leisure activities, sleeping 7–8 h a day, and staying outside the BMI ranges considered as underweight or obesity are habits that each separately has a greater probability associated with survival after a period of several years.
In all the analyses carried out, we have detected high and statistically significant levels of heterogeneity. This means that in this field of research, like any other that studies complex, high-level constructs, the levels of control over variables are low. Longevity is associated with multiple factors, both environmental and genetic, whose isolated and interacting effects have combined impacts that are very difficult to isolate. Primary studies identify factors with significant associations, but they are probably multi-moderated effects. Furthermore, the operationalization of the factors is very diverse and not always clear, in addition to being frequently based on self-reports that are not always sufficiently reliable. Let us think of the varied operationalizations that have been used of walking or diet, as well as the problems of fidelity of memory about them. The fact that significant factors appear despite these difficulties is remarkable. Nevertheless, we must not forget that this high level of heterogeneity does not allow us to accurately predict the results of future similar studies.
Regarding regular physical activity, we analyzed 61 studies which were operationalized and quantified differently. Therefore, our classification includes several possible measures used in the studies reviewed:
In our study,
Furthermore, in this meta-analysis the median of the age is 64.5 years, thus the menopause may play a central role. Before the menopause, estrogen protects the female cardiovascular system through multiple mechanisms, but after menopause, the decline in estrogen levels may be harmful (Muka et al.,
Our results show people who frequently
Physical activity has been shown to enhance the immune function mainly in less fit subjects or the sedentary population (Romeo et al.,
Regarding healthy diet, 14 studies were synthetized. Consistent with previous investigations, this meta-analysis substantiates the protective association among healthy diet and mortality. Following a healthy diet has a mortality rate 15% lower than in those reporting an unhealthy diet.
During recent decades, there has been growing research on the possibly protective role of dietary factors such as antioxidants and other micronutrients (e.g., minerals, polyphenolic compounds, phytoestrogens), generating increased research into diets rich in fruit and vegetables, under the assumption that an increase in their consumption would reduce the incidence of cancer and cardiovascular disease (Ness and Powles,
Hours of Sleep is an important factor in predicting not only the quality of sleep but also health and survival. Our results show sleeping 7–8 h/day is associated with a mortality risk of 13% less than people who normally sleep 9 h or more or 6 h or less, supporting previous research that found associations between inappropriate sleep duration and mortality, cardiovascular disease and general health in middle age (Yeo et al.,
As some authors have suggested, this association might be due to sleep deprivation causing alterations in cortisol secretion and altered growth hormone metabolism (Spiegel et al.,
In our analysis, this association is moderated by years of follow-up. The association is smaller the more years it is monitored during the study, perhaps due to changes in the sleeping pattern along the study or to other factors. Furthermore, sleep duration decreases across age (Chaput et al.,
Our findings highlight that being in the BMI ranges considered
In contrast to previous meta-analyses, our study did not find a significant association between overweight (BMI ranges 25–28) and mortality, suggesting the protective metabolic effects of increased body fat. Furthermore, Zajacova et al. (
The present meta-analysis includes large published studies representing in total more than 2,800,000 people. The analyses included study populations from Europe, North, Central and South America, Europe, United Kingdom, Asia, and Australia. However, the studies were highly heterogeneous in their methods, with different ways of operationalizing and quantifying the variables of interest. Regardless, the categories of the studies were based on those measures, trying to create a coherent and enlightening classification for each lifestyle studied. Also, results were in line with those of other meta-analyses.
Regarding the comments in the introduction referring to the methodology, and specifically the healthy lifestyles measures, our analysis does not include any subjective measures such as satisfaction or even subjective health self-reports (indirect measures). In the four lifestyles analyzed, objective measures are also available that (if analyzed) could have complemented subjective measures: In
Although all studies were adjusted for multiple potential moderators, there are likely to remain different factors, such as unrecorded changes in exposure over time given the length of follow up, possible stressful situations, important life events occurring between baseline and follow up, etc., that could substantially affect the results.
Further research should use more objective measures, eliminating the threat of bias caused by systematic differences in healthy lifestyles. In addition to better understanding the relationship between these healthy lifestyles and mortality, researchers should address the differences between: (1) studies carried out with objective measures and (2) studies with self-reported data, and analyse this categorization as a moderator. Does the effect size change, and how much? Response distortion must be controlled when self-reports are taken as measures. There must be a control of two types of distortions to self-reports (Fernández-Ballesteros and Botella,
In sum, perhaps the most critical aspect in meta-analysis is the appropriateness of the methods used by researchers in outcome evaluation and some restriction must be introduced as criteria for inclusion in health evaluation research program.
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
RF-B is the IP of the project. RF-B, MS-I, and JB designed the meta-analysis. EV-L extracted the information from the studies and performed the codification. EV-L and JB designed and performed the statistical analyses. RF-B and MS-I contributed to the interpretation of the results. RF-B, EV-L, and MS-I wrote the manuscript with support from JB. All authors contributed to the final version of the manuscript.
This study is one of the objectives of the Research Project granted by the Spanish Ministry of Science and Innovation: Project: PID2019-109761RB-I00.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
The Supplementary Material for this article can be found online at:
*Studies included in the meta-analysis are denoted in the References with an asterisk.