Article Text

Download PDFPDF

Physical activity and risk of Parkinson’s disease: a prospective cohort study
  1. G Logroscino1,
  2. H D Sesso2,
  3. R S Paffenbarger, Jr3,
  4. I-M Lee2
  1. 1Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts, USA
  2. 2Division of Preventive Medicine, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, USA
  3. 3Division of Epidemiology, Stanford University School of Medicine, Stanford, California, USA
  1. Correspondence to:
 G Logroscino
 Department of Epidemiology, Harvard School of Public Health, Boston, MA 02115, USA; glogrosc{at}


Background: Despite the fact that Parkinson’s disease is the second most common neurodegenerative disease, little is known about risk factors for the disease. Laboratory experiments indicate that physical activity may have a neuroprotective effect; however, there are few data on whether physical activity is associated with decreased risk of Parkinson’s disease.

Aim: To investigate the relationship between physical activity and Parkinson’s disease in 10 714 men (mean age, 67.6 years) from the Harvard Alumni Health Study, who were diagnosed as free of self-reported Parkinson’s disease in 1988.

Methods: Physical activity was assessed in 1988 by asking about the daily number of blocks walked and stairs climbed, and participation in sports and recreational activities in the past week. Energy expenditure was then estimated and men were categorised into four groups: <1000, 1000–1999, 2000–2999 or ⩾3000 kcal/week. In addition, physical activity data were available for three past time points: during college, 1962 or 1966, and 1977. Incident cases of Parkinson’s disease occurring after 1988 (n = 101) were identified through a follow-up health questionnaire in 1993 and death certificates obtained until 1997.

Results: In multivariate analyses, the relative risks (RR) for Parkinson’s disease associated with <1000, 1000–1999, 2000–2999 and ⩾3000 kcal/week of physical activity were 1 (referent), 1.15 (95% confidence interval (95% CI) 0.71 to 1.88), 0.92 (0.50 to 1.71) and 0.63 (0.36 to 1.12), respectively; p for trend was 0.12. When walking was examined separately, somewhat lower, but not significant, risks were observed for Parkinson’s disease. The multivariate RRs (95% CI) for walking <5, 5–10, 10–20 and >20 km/week were 1 (referent), 0.67 (0.37 to 1.23), 0.81 (0.50 to 1.31) and 0.72 (0.39 to 1.34), respectively; p for trend was 0.26. Analyses that considered physical activity at other time points before 1988 did not show any significant associations.

Conclusions: These data do not strongly support the hypothesis that physical activity lowers the risk of Parkinson’s disease. However, as the number of patients with Parkinson’s disease in this study was not large, statistical power may have been limited and further large studies are needed to provide additional data.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Although Parkinson’s disease is the second most common neurodegenerative disease among those aged >65 years, the aetiology of Parkinson’s disease is not well understood. Only a few risk factors have been consistently identified, with both smoking and caffeine showing strong inverse associations with the disease1 and family history showing a positive association.2 Other risk factors such as specific foods or nutrients have been identified in some studies, but not in others.3–7

A potentially interesting and modifiable risk factor for Parkinson’s disease is physical activity. Physical activity has been shown to be inversely related to other neurodegenerative diseases such as Alzheimer’s disease and cognitive decline.8,9 Animal models of Parkinson’s disease show that treadmill training resulted in the attenuation of dopamine depletion in the striatum compared with that in control animals.10 In other experiments, exercise training improved both behavioural and motor symptoms.11 In humans, four previous studies have investigated the role of physical activity in preventing Parkinson’s disease.12–15 In the only prospective study, men who exercised strenuously had a 60% decreased risk of Parkinson’s disease compared with men who did not, but no clear effect was observed in women.15 To provide additional prospective data, we evaluated the relationship between physical activity and risk of Parkinson’s disease in an ongoing study, the Harvard Alumni Health Study.



The Harvard Alumni Health Study is an ongoing follow-up study of men who entered Harvard University as undergraduate students between 1916 and 1950. The cohort was established when 21 582 alumni responded to mailed questionnaires in 1962 or 1966, with questions about demographics, medical history and health practices. Subsequently, men were contacted at periodic intervals to update the information provided. In all, 12 805 men responding to the 1988 questionnaire were eligible for this analysis. We excluded men who reported Parkinson’s disease on this questionnaire (n = 36) or who did not provide information on the disease (n = 2). We also excluded men with missing information on physical activity (n = 505). Of the remaining 12 262 men, we successfully followed up 10 714 (87%) men.

Assessment of physical activity

On the 1988 questionnaire, men reported their daily walking and stair climbing activities, as well as sports and recreational activities performed in the previous week.16 This assessment of physical activity has proved to be reliable and valid.17–19 For example, the test–retest correlation coefficient over 1 month was 0.72, whereas estimates of energy expenditure from questionnaires compared with physical activity records yielded a correlation coefficient of 0.65.19 On the basis of the information provided, we estimated the energy expended on physical activity in kilocalories per week, and divided men into four groups: <1000, 1000–1999, 2000–2999 and ⩾3000 kcal/week. In addition to the 1988 data, identical information on physical activity was collected from another questionnaire in 1962 or 1966, as well as one in 1977 for 7858 men. Further, information on sports played during college was available for 9501 men (categorised as intramural sports activity of <5 h/week, ⩾5 h/week or of unknown time per week, or varsity athlete (ie, representing the university in sports competition)).

Assessment of potential confounders

On the 1988 questionnaire, we also obtained information on cigarette smoking (categorised as never, past or current smoker), tea and coffee intake (for each, categorised as almost never, up to 1, 1–2 or >2 cups/day) and whether men had been diagnosed to have cardiovascular disease and cancer (no or yes).

Ascertainment of Parkinson’s disease

On a subsequent health questionnaire mailed in 1993, we asked men about doctor-diagnosed Parkinson’s disease. Men were also followed up for mortality until 1997, and death certificates were obtained. Using both methods, we identified 101 men with Parkinson’s disease during follow-up. Self-reported Parkinson’s disease in this cohort is reasonably valid: in a previous validation study, self-reported diagnosis of Parkinson’s disease was confirmed in all patients by the treating physician (70.1% of doctors agreed to participate in the validation study).14

Statistical analyses

We examined the baseline characteristics of the subjects according to the different levels of physical activity, and tested for differences using the χ2 test for categorical variables and the t test for continuous variables. We used Cox’s proportional hazards models to estimate the relative risks (RRs) of Parkinson’s disease associated with different levels of physical activity. There was no indication that the proportional hazards assumption was violated. We initially adjusted the estimates only for age. We then also controlled for other known risk factors for Parkinson’s disease: smoking, and tea and coffee consumption. Additionally, we adjusted for previous diagnosis of cardiovascular disease or cancer, as these chronic diseases may have altered men’s physical activity levels. In further analyses, we adjusted for alcohol intake and body mass index, which have been inconsistently related to Parkinson’s disease risk. However, the results were unchanged, and so we present only results that are adjusted only for age, smoking, tea and coffee consumption, and previous diagnosis of cardiovascular disease and cancer.


Table 1 shows the baseline characteristics of men according to their physical activity in 1988. The mean age of all 10 714 men at baseline was 67.6 (standard deviation 8.1) years. Men who were more active were younger, less likely to smoke cigarettes and more likely to drink tea at baseline. They were also less likely to report a history of cardiovascular disease or cancer. We then compared the characteristics of men included in this study with those lost to follow-up. The two groups of men were similar with regard to the characteristics in table 1, including physical activity (data not shown).

Table 1

 Baseline characteristics by physical activity level in 1988, Harvard Alumni Health Study

During 61 644 person-years of follow-up, we identified 101 of 10 114 men with Parkinson’s disease. This observed incidence of Parkinson’s disease is similar to that previously reported among Caucasian men, in whom diagnosis of Parkinson’s disease was based on clinical examination, supporting the validity of our case ascertainment.20,21 In age-adjusted analysis, we found no significant association between the total energy expended on physical activity in 1988 and risk of Parkinson’s disease (table 2), with a non-significant reduction in risk at ⩾3000 kcal/week compared with <1000 kcal/week (RR 0.64; 95% confidence interval (CI) 0.36 to 1.13). When we examined the different activity components, we also found no significant associations with walking, stair climbing or participating in sports or recreation of light, moderate or vigorous intensities (RR = 0.67; 95% CI 0.37 to 1.23; p for trend = 0.26). There was a suggestion of lower risk, although non-significant, associated with walking 5 to <10 km/week compared with the risk with walking <5 km/week, with no significant risk reduction at higher levels of walking. Further adjustment for smoking, tea, coffee and alcohol consumption, and history of cardiovascular disease or cancer did not appreciably change the findings.

Table 2

 Relative risks of Parkinson’s disease, 1988–97, according to physical activity in 1988, Harvard Alumni Health Study

We then investigated whether long-term physical activity influenced the risk of Parkinson’s disease. A total of 7858 men provided information on physical activity in 1962 or 1966, 1977 and 1988. To estimate the long-term physical activity, we averaged energy expenditure over these three years. Table 3 shows that higher average levels of long-term physical activity did not predict the risk of Parkinson’s disease developing after 1988 among these 7858 men (multivariate p for trend 0.98). We also examined physical activity in 1962 or 1966 and in 1977, separately, in relation to risk of Parkinson’s disease, and still did not find any significant associations (p for trend 0.24 and 0.70, respectively). Finally, we investigated physical activity in the remote past, during college (with data available for 9501 men), in relation to risk of Parkinson’s disease (table 4). Varsity athletes, who exercised at a much higher level than men playing only intramural sports for <5 h/week, did not experience a lower risk (multivariate RR 1.09; 95% CI 0.41 to 2.93).

Table 3

 Relative risks of Parkinson’s disease, 1988–97, according to long-term physical activity, Harvard Alumni Health Study

Table 4

 Relative risks of Parkinson’s disease, 1988–97, according to physical activity during college, Harvard Alumni Health Study


The results of the present study do not provide strong support for the hypothesis that physical activity protects against Parkinson’s disease in men. Although there was a suggestion of reduced risk among men expending ⩾3000 kcal/week in physical activity at baseline (1988), as well as with greater distances walked, none of these relative risks was statistically significant. Additionally, there was no indication of a lag effect, in that physical activity either in the distant past or averaged over the long-term past did not affect the risk of Parkinson’s disease. However, the number of cases of Parkinson’s disease was not large, and statistical power may have been limited.

The present study is an updated analysis of the association between physical activity and risk of Parkinson’s disease in the Harvard Alumni Health Study. A previous nested case–control analysis of cases diagnosed before 1988 also suggested inverse, non-significant associations with physical activity.14 There are few other studies of physical activity and risk of Parkinson’s disease. Of two case–control studies12,13 examining the association, only one found a significant inverse relationship with physical activity, but this finding was limited only to people with early-onset Parkinson’s disease.12 Most previous studies of this topic have several potential methodological problems: only one was a prospective study with a nested case–control design,14 whereas two others included prevalent cases with retrospective assessment of physical activity12,13 and small sample size.13 Recently, the findings of a prospective cohort study were published, showing an inverse association with physical activity in men but not in women.15 The authors suggested that limited power, due to the few cases of Parkinson’s disease among women, was the most likely explanation for the difference in effect between the sexes.

Biologically, it is plausible for physical activity to decrease the risk of developing neurodegenerative diseases. Several mechanisms may be responsible for the protective effect of physical activity on the brain, such as improvement in cerebrovascular circulation22 and increased production of several neurotransmitters, including neurotrophic substances.23 Some studies indicate that exercise may also have a role in decreasing age-dependent neuronal loss and age-related decline in spontaneous motor activity.23–25 Physical activity may protect against risk of Parkinson’s disease via some of these neuroprotective mechanisms. Additionally, physical activity may protect against Parkinson’s disease by directly affecting dopamine production. In one animal experiment, dopamine levels in the neurostriatum increased by about 30% in mice forced to run.26

In epidemiological studies, factors associated with a westernised lifestyle, including lack of exercise, may contribute to the increased prevalence of neurodegenerative diseases.27 In a prospective cohort study conducted in four clinical centres in the US, elderly women who were more physically active had a 40% reduced risk of cognitive decline compared with sedentary women.9 In a community-based study, the Canadian study of Health and Aging, high levels of physical activity halved the risk of onset of Alzheimer’s dementia and dementia of any type.8,9,28

Some limitations should be considered in the interpretation of our results. Firstly, we ascertained Parkinson’s disease using self-reports and through death certificates. However, we do not believe this to be a major limitation. A previous validation study showed a high concordance between self-reports and doctors’ diagnosis.14 In a recent study conducted on the general population, 78.1% of people who self-reported Parkinson’s disease also had a medical insurance claim with a Parkinson’s disease diagnosis, filed by their doctors.30 Participants in the present study, who included college-educated men of higher socioeconomic status, would therefore be expected to report a diagnosis of Parkinson’s disease with reasonable validity. As for the use of death certificates, a recent study indicated that the quality of death certificates for ascertaining Parkinson’s disease is high among well-educated people.29 Further, the incidence rates of Parkinson’s disease across all age groups in the present study are similar to those in population-based studies in Europe and the US, where clinical diagnosis was used to ascertain Parkinson’s disease.20,21 We acknowledge that other forms of parkinsonism, such as progressive supranuclear palsy and multisystem atrophy, may have been incorrectly diagnosed as idiopathic Parkinson’s disease; however, the prevalence of neurodegenerative parkinsonism other than idiopathic Parkinson’s disease is very low in the general population.31

A second limitation is that it is unclear how long the preclinical period of Parkinson’s disease may be, and the time window between the relevant period of exposure and disease onset may be many years.32 In analysing the association of physical activity in 1988 with risk of Parkinson’s disease between 1988 and 1997, some of the people with Parkinson’s disease may already have been in the preclinical phase in 1988 when we ascertained physical activity and they might, consequently, have decreased their physical activity. As recently reported,33 the decline in physical activity may precede the clinical onset of the disease by as many as 8 years. However, when we analysed physical activity at different time periods in the remote past, we also observed no significant associations. We also need to consider the possibility that some people may have limited their physical activity in later life because of early symptoms of age-related diseases or poor general health. We dealt with this concern by adjusting for prevalent cardiovascular disease and cancer in 1988; however, we cannot exclude the effect of other unaccounted for chronic conditions.

A third concern is that people with a high level of physical activity may be more health conscious and more likely to seek medical care for early symptoms of Parkinson’s disease. However, in this group of men with high socioeconomic status and access to medical care, we believe that any bias from selective diagnosis among active men would be small. Finally, participants in the present study were well educated, of high socioeconomic status and primarily white people, potentially limiting the generalisability of the study. However, the possible underlying mechanisms between physical activity and Parkinson’s disease risk probably do not differ for other population groups.

Strengths of the present study include the prospective design with a high follow-up rate, reducing the likelihood of selection bias in cohort studies. We also had detailed information about time of physical activity, which, although self-reported, has good validity and reliability.17–19 Finally, we had detailed information on the health habits and health history of participants, and so were able to control for important potential confounders previously reported in the literature.

In summary, although we observed some non-significant inverse associations, the present study does not strongly support a beneficial effect of physical activity on risk of Parkinson’s disease despite plausible biological mechanisms. Further large prospective studies, investigating different types of physical activity over the whole life span in relation to risk of Parkinson’s disease risk, are needed to provide additional data. Confirmation of any beneficial effect of physical activity on Parkinson’s disease in such studies would have an important bearing on public health.


We thank Sarah E Freeman, Rita W Leung, Doris C Rosoff and Alvin L Wing for their help with the College Alumni Health Study.



  • Published Online First 22 August 2006

  • Funding: This research was supported by grant HL077 548 from the National Heart, Lung, and Blood Institute.

  • Competing interests: None declared.

  • This is report no LXXXVII in a series on chronic disease in former college students.