An increase in the proportion of elderly in the population has been reported in developed [1, 2] and developing countries [3–6]. The number of people over 60 years old is projected to double in the next 20 years. According to demographic projections, 33 million Brazilians will be older than 60 years in 2025 [1]. Furthermore, the prevalence of physical inactivity is increasing worldwide as a result of changes in individuals' lifestyles. Studies conducted in Brazil reported that nearly 45.0% of the people [7, 8] and 50.1% of the elderly are sedentary [8]. Data from a national survey called VIGITEL (" "Surveillance of Risk and Protective Factors for Chronic Diseases by Telephone Interviews") in all capitals of the 26 Brazilian states found that only 12.7% of the elderly were involved in physical activity in leisure time, whereas 56.5% were classified as physically inactive [9]. A recent report from the World Health Organization (WHO) demonstrates that physical inactivity is the fourth-leading global risk factor for mortality [10] and is one of the most important modifiable risk factors for noncommunicable diseases (NCDs), such as heart disease, diabetes and cancer, contributing substantially to the global burden of disease, disability and death. Physical activity can prevent and help treat NCDs and maintain physical and mental health and quality of life in elderly adults [10–12]. Moreover, low levels of physical activity have been associated with high health care costs [13]. Therefore, national data on the prevalence of and factors related to physical activity level may aid in identifying effective public health measures, consequently lowering health care charges.

Population studies have been conducted in developed and developing countries on physical activity and health-related factors among the elderly [14–18]. However, few published studies have specifically investigated physical activity level determinants, among older adults in Brazil [7, 8, 19, 20].

The Epidoso ("Epidemiologia do envelhecimento" - "Epidemiology of aging") Project was the first longitudinal study of the elderly in São Paulo, a large urban center in Brazil. This study has followed elderly people since 1990, searching for factors associated with healthy aging and risk factors for mortality. This project has resulted in important contributions, through some papers in the area of risk factors for mortality, functional capacity, falls and noncommunicable diseases, as well as the prevalence of several diseases in this population [4, 5, 21]. Therefore, we used data from the Epidoso Project to identify health-related factors that independently influence behavior trends in physical activity level over the course of two years among the elderly. As far as we know, no study has been conducted with the Brazilian elderly that underlines factors that influence trends in PAL over two years.

Over the last decade, interest in the impact of PAL on health among the elderly population has increased. Physical inactivity is common among elderly and this finding is particularly alarming considering that this population is greatly affected by noncommunicable diseases (NCDs). The current study found significant associations between behavior trends in physical activity level and age, gender, ADL scores and falls in a Brazilian urban community elderly population living in São Paulo state.

The prevalence of physical inactivity (87%) found in the baseline data was higher than that in another city in the south of Brazil (50.1%) [8] and in national data (56.5%) [9]. In addition, the prevalence of physical inactivity in leisure time among older adults was comparable to that of a large city in Vietnam (91%), but higher than that in four Lebanese districts (30%) [14, 15]. In the U.S more than 30% of the adults above 70 years old were considered inactive in 2004 [32]. A possible explanation for this difference may be varying definitions of PA. The current study evaluated total PA considering only activities in leisure time. Some epidemiological studies in Brazil have demonstrated the importance of domestic and routines activities as part of the total PAL in the elderly population [7, 8].

After two years, a decline in physical inactivity was observed in the elderly. However, when the data were evaluated more accurately, most of the elderly (84.4%) experienced a not favorable behavior trend in PAL, which included the participants who remained physically inactive and those who displayed decreased PAL (Insufficiently Active to Physically Inactive and Regularly Active to Insufficiently Active). In addition, only 15.6% (n = 134) remained active/insufficiently active or increased their PAL (Physically inactive → Insufficiently Active; Insufficiently Active → Regularly Active). Our finding is in line with other studies [32]. A study conducted in the U.S. showed a decline in physical inactivity (from 30% in 1994 to 24% in 2004) mainly in the age groups of 50 to 59 and 60 to 69 years old [32]. These findings suggest that despite the recognized benefits of PA for physical and psychological health and mortality [10–12, 33], many older adults remain physically inactive and maintain this negative behavior. Additionally, considering that initiating activities in leisure time even in midlife increases the probability of successful survival or exceptional overall health in later life [33], public health efforts should emphasize the promotion of physical activity and identify the factors that limit physical activities among older adults.

Considering age, the present data are in agreement with information presented in the literature [7, 8, 14–17, 19, 34], which shows a gradual decrease in physical activity participation with an increase in age. In Poland, for example, the risk of leisure time physical inactivity was significantly higher among persons older than 64 years of age. Our study also demonstrated that the oldest elderly have a lesser chance of becoming physically active or insufficiently active despite their initial PAL, which may be attributable to a high proportion of the elderly requiring assistance in activities of daily living and to a decrease in work-related activities [34].

In the present study, men were more active than women and improved their PAL after two years. Similar findings have been extensively reported [7, 16, 20, 34, 35]; however, most of them, including our research, did not evaluate household activities, which contribute considerably to the overall physical activity level among elderly women. A study conducted in Vietnam observed that elderly women were more active than men [15]. This difference could be explained by the fact that a high proportion of women are involved in household activities even in later life, contributing to the overall PAL [15].

There is strong evidence supporting an association between educational level and physical activity in the elderly [7, 14–16, 36, 37]. Our study demonstrated that elderly individuals with a poor education were more physically inactive in leisure time. Furthermore, after 2 years, they became physically inactive or insufficiently inactive despite their initial physical activity level. There are several possible explanations for this result. First, those with a lower level of education might have a lack of knowledge of PA benefits and fewer opportunities to participate in physical activities during leisure time. Second, in developing countries, occupation/work activities are important components of total activity level in addition to leisure-time activities, even among the elderly [14, 16].

The relationship between functional capacity and PAL has been demonstrated in several studies [19, 38–40]. Data from "The Women's Health and Aging Study" have demonstrated that physical inactivity is more frequent among women with a lower functional capacity [39]. A similar result is seen in the south of Brazil. Moderate to severe difficulties in performing activities of daily living were associated to physical inactivity among elderly women [19]. These results are in agreement with the results from the current study, which indicates that more than four limitations in ADL score is positively associated with less physical activity. Furthermore, after two years, the prevalence rates of becoming physically inactive increased with rising difficulties in ADL. Therefore, there are strong evidences showing the relationship of exercise with functional capacity [19, 38–40].

A strong association was found between physical inactivity and the Mini-Mental State Examination in the first analysis. The majority of subjects that experienced a decrease in PAL after two years presented lower values for cognitive status. However, when controlled by other variables, this aspect did not show a correlation with PAL. These results are contrary to findings from other countries in which low levels of physical activity were associated with poorer cognitive performance [41, 42]. One of the analyzes of the Duke Longitudinal Study, Texas, U.S., showed lower scores in cognitive tests in the elderly group that retired and remained sedentary compared with the group that retired but continued performing regular physical activities. Furthermore, those that retired and later adopted a sedentary lifestyle presented a greater risk of brain-vascular disease and possible cognitive decline compared to the group that retired but continued performing regular physical activities [41]. Among 18,766 elderly nurses between 70 and 81 years old in the United States, long-term regular physical activity, including walking, was associated with significantly better cognitive function and less cognitive decline [43]. However, the inverse relationship has not been well established in prospective studies [44].

No strong associations were found between physical inactivity and BMI or the dysthymia screen. Likewise, no significant relationship was reported between these variables and behavior trends in PAL over two years.

Some cross-sectional studies found significant association between PAL and BMI [14, 17]; nevertheless, the present results are in agreement with other Brazilian studies [8, 45]. Despite the benefits obtained by exercise (an increase in rest metabolic rate and a decrease in risk of cardiovascular diseases), there are no data that show a change in adiposity or in body weight of the elderly with an exercise program that does not include dietary restriction [46]. This result may explain the findings of our study, including the fact that we did not find changes in the BMI. However, we cannot prove this hypothesis because in our study, dietary habits and other variables that may influence body composition were not considered longitudinally.

A positive relationship between physical inactivity and psychological well-being in the elderly without a clinical disorder has been reported in meta-analysis studies [47]. However, data from other studies have demonstrated conflicting results [47–49]. A recent study showed that the elderly with emerging depression in Amsterdam, the Netherlands, tended to change more often to a sedentary lifestyle than those without depression, although no causal relationship was demonstrated [49]. Because depression is closely related to functional decline in older adults [50], when ADL were controlled in the final logistic regression model, no strong association was found between scores of dysthymia screen and behavior trends in PAL. Additionally, response bias and cohort differences are problems related to self-report, and perceptions of well-being differ in older cohorts [51]. Furthermore, most of the findings are from studies that have evaluated the influence and benefits of an active lifestyle for depression and among younger groups.

Our finding of correlation between falls and a not favorable behavior trends in PAL is in line with other studies [52–54]. In a cross-sectional study Zijlstra and colleagues [53] verified that multiple falls were independently associated (OR = 4.64; 95% CI = 3.73-5.76) with the avoidance of activity among 4,031 community-living elderly. A study conducted in Belgium demonstrated that falls were correlated with the avoidance of activities in community-living elderly between 61 and 92 years old, both in cross-sectional and in longitudinal analyses [54]. An interesting hypothesis for these findings may be the gradual development of avoidance behavior at the mobility level, which results in a deterioration of physical abilities and falls at home in the long term. Previous cross-sectional studies have shown a correlation between falls and fear of falling, but it is unclear which occurs first. Several studies [53, 54] have pointed out that fear-related avoidance of activities may have negative effects on physical abilities and may also be predictive of future falls. Falls are a common and important problem in older community-dwelling adults. One of the major consequences of falls is avoidance activities, which consequently decrease PAL. Future research should focus on strategies to prevent falls within this population.

Although we did not observe an independent effect of fractures on PAL behavior trends, some studies have shown the importance of fractures in decreasing PAL [55, 56]. Despite the potential benefits associated with exercise after a fracture, elderly from Baltimore Hip Studies who sustained hip fractures were not likely to engage in regular exercise (resistive or aerobic). Furthermore, many factors may influence older adults' motivation and willingness to exercise, such as poor health, disability, lack of knowledge about exercise and its benefits, lower sense of self-efficacy for exercise [56], and fear of falling and injury [53, 54].

Noncommunicable diseases were not analyzed in the present study but may be associated with and/or explain changes in some variables after a follow-up period.

Physical activity involvement, mainly the increase in physical activity level and the maintenance of an active/insufficient level, must be emphasized as a health-promotion intervention policy for this group.

The limitations of our study should be considered when interpreting the results. First, the lack of a question regarding the intensity of physical activities must be considered. However, for the most part, studies have shown that elderly people tend to adopt physical activities of moderate intensity. Additionally, we did not consider or measure the influence of other variables (confounders) that may be related to body mass index, functional capacity, cognitive and dysthymia screen, falls and fractures (statistical analysis). Another limit was the short follow-up period (two years). Furthermore, we did not analyze other components of overall physical activity level. Individuals who are inactive during leisure time may be more active in other contexts. In developing countries, household and routine activities contribute greatly to total PA. In addition, because other studies used different definitions of physical inactivity, instruments and cut-off points, comparisons are difficult. Finally, the difference of almost twenty years between the first stage of the Project and the publication of the results is not considerable, while our main purpose was to verify the relationship between health-related factors and PAL trends.

This study demonstrated that gender, age, dependence in ADL and falls (after adjustment for some potential confounders) were associated with PAL and its trends over two years. Being older and female with poor functional capacity and having experienced falls contributes to a not favorable behavior trend in PAL among the elderly longitudinally. Our findings highlight the need for health-promotion programs that target these factors related to PAL, reducing barriers to achieve desired changes. The development and implementation of programs to promote a physically active lifestyle must be encouraged, even in later phases of life.