Upper respiratory tract infection (URTI), including common cold and influenza, is the most common reason for seeking primary care in many countries (12). Despite that, little is known about the potential strategies to reduce susceptibility.
Several lines of research indicate that physical activity is beneficial for a person's health. Moderate physical activity has been associated with reduced risks of a large number of diseases, such as type 2 diabetes (15), cardiovascular diseases (36), and certain cancers (41), and on overall mortality (37). Moderate physical activity has also been associated with reduced inflammation and has been suggested to boost immune responses to pathogens (19,27). A J-shaped association between physical activity and URTI was suggested in 1994 by Nieman (24), in which moderate levels of physical activity are associated with a reduced risk of URTI, whereas very high levels of physical activity, such as for marathon runners and elite cyclists, increase susceptibility to URTI. The majority of previous studies on physical activity and URTI have focused on athletes, and many have been conducted with small study populations, with fewer than 40 individuals (for a review, see Nieman [25]). Consequently, little is known about the effects of moderate levels of physical activity on URTI. A cohort study of 20- to 70-yr-old men and women (N = 547) found that moderate to vigorous physical activity was associated with a 20% lower risk of self-reported URTI (20). Another cohort study of 50- to 69-yr-old male smokers (N = 14,400) reported a weak association near null (RR = 0.93, 95% confidence interval (CI) = 0.85-1.02) between physical activity and URTI (14).
Chronic stress is associated with suppression of both cellular and humoral measures of immune function (32), and self-reported perceived stress over the last month is associated with an increased risk of URTI (for a review, see Cohen [5rsqb;). Physically active subjects show attenuated cardiovascular and stress hormonal responses to mental stress as compared with physically inactive subjects (9). Given the corroboration from experimental animal models (11), it has been hypothesized that physical activity might protect against the immunosuppressive effects of chronic stress (8).
In this study, we focus on the relationship between physical activity and self-reported URTI and examined the possible interactive relationship between physical activity, perceived stress, and self-reported URTI. The effects on immune function of perceived stress have been well studied, as have the effects of intense physical activity in athletes. In contrast, little is known about the effect of physical activity among nonathletes and the potential interaction between physical activity and perceived stress on immune function.
Because URTI typically has a brief duration, frequent follow-ups are needed for an accurate incidence tally. In this study, we relied on e-mail to prompt responses for short Web-based questionnaires to assess incidence of respiratory infections. In a previous study, we found that Web questionnaires generated good response rates for follow-ups in comparison with mailed questionnaires (4).
SUBJECTS AND METHODS
The Lifestyle and Immune Function (LIME) study is a population-based prospective cohort study of lifestyle factors and immune function, using the Internet as a novel tool for data collection. In January 2004, we selected participants randomly from the Swedish Population Registry and invited them to participate in the study. The study was approved by the ethics committee at the Karolinska Institutet, and filling out the questionnaire was considered as informed written consent. The study population comprised of 1509 men and women, aged 20-60 yr, residing in a middle-sized county in Sweden with a level of urbanicity (80%) typical for Sweden.
Study design.
Invitations to participate in the study were sent out via regular article mail. The invitations included the information on how to access the Web questionnaire, the details on the use of a Web browser, the URL to our Web questionnaire, and an individual username. The baseline questionnaire about lifestyle factors included a question about the participant's e-mail address. We sent five follow-up questionnaires during the following 15 wk (in February, March, early April, late April, and May). Every follow-up questionnaire included questions on URTI during the three preceding weeks. We sent reminders to nonresponders by e-mail 1.5 wk after each follow-up questionnaire.
The design of the study is illustrated in Figure 1. In total, 5000 individuals were invited to participate in the study, 1805 completed the baseline Web questionnaire, and 1509 were eligible for follow-up questionnaires after exclusions. We excluded participants who had URTI at baseline (n = 236), lacked an e-mail address (n = 17), or chose not to disclose it (n = 20). During the study, it became evident to us that participants at one specific workplace had an e-mail server that filtered our e-mails as junk mail. We excluded these participants from the study (n = 23) because they could not be invited via e-mail to fill out the follow-up questionnaires. Response rate for each follow-up questionnaire ranged between 83% and 84% (number of responders to that follow-up questionnaire divided by the number eligible for that follow-up questionnaire). In total, 1111 responded (74% of baseline respondents) to all five follow-ups.
;)
FIGURE 1: Description of study design and response rates, the LIME study.
The baseline Web questionnaire has been described in detail elsewhere (4). In brief, the questionnaire included automatic checks for incomplete or implausible answers, reminder messages to the respondent when a question was left unanswered, hiding of nonrelevant follow-up questions, automatic summarization of answers, voluntary personalized feedback, and illustrations to clarify complex questions. No software installation was required to complete the Web questionnaire.
Exposure assessment.
Total physical activity and inactivity was assessed at baseline. Participants were asked to estimate the amount of time (hours and minutes) spent on each out of nine activity levels, including sleep, adding up to 24 h, for a usual day and night. Each activity level was explained by examples of activities corresponding to a MET value (MET, MET task, multiples of resting metabolic rate) (2,3). We estimated average MET-hours expended during a 24-h period by multiplying MET level by the time spent at each activity level. The MET values assigned to the nine different levels ranged between 0.9 and 8.0 METs. One MET is equivalent to sitting in a relaxed position. Activities equal to 3-5 METs include walking at a moderate to brisk pace, housework, or playing golf, whereas mowing the lawn with a hand mower or jogging at a slow pace is equal to 6 METs (2,3). Hours and minutes spent in different activities were automatically summed in the Web questionnaire, and participants were promptly reminded if the total hours and minutes for a typical day did not add up to 24 h. The method was developed and validated by Lagerros et al. (17).
On the basis of a cross-sectional study using a Danish version of the physical activity assessment method (1), we categorized total physical activity (MET-hours) into three levels (<45, 45 to <55, and ≥55 MET·h·d−1), with the lowest group used as reference. In addition, we defined nonvigorous physical activity as activities corresponding to 3-5 METs and vigorous physical activity as activities corresponding to 6 METs or more.
At baseline, we measured stress during the previous month by the Perceived Stress Scale (7), which is a set of 14 questions of general nature that measures the degree to which life situations are appraised as stressful by the participant. We defined low stress level as having ≤23 points (below median) on the basis of the Perceived Stress Scale and high stress level as having >23 points (above median). We assessed diet at baseline by a validated semiquantitative food frequency questionnaire including 96 food items (21,22).
Ascertainment of URTI.
Self-reported URTI was assessed at baseline and in all five follow-up questionnaires. In the follow-up questionnaires, participants were asked if they currently had an infection (cold or influenza) or if they had had a new infection during the last 3 wk or since the last questionnaire. Participants were considered to have a URTI if they answered "yes" to this question. They were instructed not to count a URTI episode twice, even if it crossed over two follow-up periods. Follow-up questions about symptoms were given to all participants who reported an infection. Seven symptoms were recorded: sore throat, cough, runny nose, headache, malaise, fever, and unspecified symptoms. This information was used to classify cases into systemic and nonsystemic URTI. Cases reporting symptoms of fever were classified as systemic URTI, whereas cases without fever were considered nonsystemic URTI.
The influenza season of 2003-2004 in Sweden was of medium intensity. Activity peaked during the last week of December 2003 and the first week of January 2004 but declined shortly after and remained low during the rest of the season, which ended in the middle of March (18). Most (94%) of confirmed cases were of the influenza A H3 strain (18). Influenza vaccination history was collected at baseline and updated at every follow-up. Information on pollen allergy was collected in the follow-up questionnaires in April and May (the pollen season in the part of the country where the study population lived) because symptoms of pollen allergy can mimic a URTI. In total, there were 1181 self-reported cases of URTI in a total of 16,984.5 person-weeks of follow-up for URTI.
Data analysis methods
We divided the number of reported URTI by person-time at risk to get incidence rates. We standardized for age and sex using the age-sex person-time distribution of the low physical activity group as the standard (30). We also estimated incidence rate ratios (IRR) with 95% CI using Poisson regression models to assess and to control age and sex along with other confounding factors. Disease-free participants contributed 3 wk of time under risk (i.e., person-time) for each 3-wk follow-up period. A participant with no reported URTI could therefore contribute up to five 3-wk periods of time under risk, a total of 15 person-weeks. Because we did not know exactly when a URTI occurred or how long it lasted during the 3-wk period, we assigned participants who reported a URTI 1.5 wk of risk time out of the 3-wk follow-up period. We considered each follow-up period to be independent of previous or later periods for the same person. The correlation in the data because of repeated measurements on the same individual was investigated using the generalizing estimating equations (38). We found no evidence of substantial correlation; therefore, the generalizing estimating equation was not used in the final analyses. On the basis of this, we assumed that the risk of contracting a URTI was independent of one's previous history of URTI.
We assessed the possibility of confounding of a large number of potential confounding factors by eliminating one factor at a time from the full model. We included in the final Poisson model those that when dropped changed the effect estimate more than 10%. These factors were age, asthma, self-reported weakened immune system, regular contact with children under age 10 yr at home or at work, regular contact with large crowds at leisure time as well as energy-adjusted (39) intakes of vitamin D, vitamin E, and selenium. Sex, body mass index, perceived stress, education level, and month had small confounding effects but were kept in the final multivariable model nonetheless. Individuals with missing information on any covariate in the final multivariable models were excluded from the analyses (n = 91 cases). We also assessed as potential confounders smoking, snuff dipping, alcohol intake, use of public transport, nursing sick people at work, contact with large crowds at work, pollen allergy, influenza vaccination history, and intake of 11 energy-adjusted nutrients, including vitamin C and zinc. None of these were an important confounder, and we omitted these variables from our final model.
The multivariable analysis of physical activity and risk of URTI included 15,780 person-weeks contributed by 1255 individuals, generating 5805 follow-up periods and 1090 cases.
We evaluated effect measure modification between physical activity and perceived stress by plotting rates for high versus low levels of perceived stress as shown in Figure 2. We fit a cubic spline model with six knots (at minimum and maximum values and at the four quintile cutoffs) (31) to ascertain the shape of the dose-response relationship, and we present the predicted rates (at baseline of all other covariates) from the model graphically.
FIGURE 2: Smoothed incidence rates of self-reported URTI for physical activity by level of stress, for all (A), for men (B), and for women (C). The smoothed rates of URTI for high and low stress over varying levels of MET-hours per day (MET task) are from a spline model. The rate is predicted at the reference level for all other covariates in the model; young (age = 20-29 yr), women (for panel A), with normal body mass index, no asthma, no weakened immune system, no contact with children at home or work, no regular contact with large crowds at leisure time, low-medium intake of vitamin D, vitamin E, second lowest selenium intake categories after adjusted for energy intake, and February as the reference point for monthly effects. The graph has been truncated to show only values between 33 and 67 MET·h·d−1.
We used Stata 10.1 Intercooled (StataCorp LP, College Station, TX) and SAS v 9.2 (SAS, Cary, NC) for the analyses.
RESULTS
Table 1 shows baseline characteristics of the participants in the LIME study by level of physical activity. We found only small differences for age, body mass index, smoking, asthma, contact with children, poor sleep, perceived stress, energy intake, and self-reported weakened immune system. Women and those with higher levels of education were more likely to report lower levels of physical activity. There were more infections during the winter than that in the spring: 23% of the participants reported a URTI in February compared with 14% in May.
TABLE 1: Baseline characteristics of participants by levels of physical activity expressed as MET-hours per day (the LIME study, Sweden 2004).
Table 2 shows the incidence rates of URTI by age, sex, and level of physical activity. In general, the incidence rate for URTI decreased with increasing age for both men and women. The age- and sex-standardized IRR of URTI by level of physical activity are shown in Table 2 and indicate that high levels of physical activity are associated with a reduced IRR of URTI among both men and women in most age groups.
TABLE 2: Incidence rates of URTI by age, sex, and level of physical activity (IRR are standardized for age and sex).
Table 3 shows IRR for physical activity from the Poisson model adjusting for age, sex, and other confounders. High levels of total physical activity (≥55 MET·h·d−1) were associated with an 18% reduced risk of URTI compared with low levels of physical activity (IRR = 0.82, 95% CI = 0.69-0.98). For nonvigorous physical activity, such as walking or doing housework, we found an inverse association (IRR = 0.89, 95% CI = 0.79-1.01) that was slightly weaker but analogous to that found for vigorous physical activity, equivalent to jogging (IRR = 0.86, 95% CI = 0.75-0.98).
TABLE 3: Physical activity and IRR of URTI.
In our study, URTI includes both common cold and influenza. We cannot with confidence separate common cold from influenza on the basis of self-reported symptoms, but to get an indication, we divided all cases into URTI cases with systemic symptoms such as fever (i.e., sign of influenza) and nonsystemic URTI cases without fever (sign of common cold) and found that the strong protective effect of physical activity remained for nonsystemic URTI, which constitutes the majority of the cases (76%). We saw little effect of physical activity for URTI with systemic symptoms (Table 3).
To rule out the influence of repeated events in the same person, the analysis was limited to only include person-time until the first URTI occurred for each person. Results were similar to findings from the main analysis: high levels of physical activity (>55 MET·h·d−1) were associated with a reduced risk of URTI compared with low levels of physical activity (IRR = 0.79, 95% CI = 0.63-0.98).
We saw no effect on the risk of URTI for perceived stress in the whole group (age- and sex-adjusted IRR = 1.05, 95% CI = 0.94-1.18). Table 4 shows IRR for physical activity by levels of perceived stress. These results indicate that the inverse association of physical activity on URTI risk was stronger in the group with high perceived stress (IRR = 0.58, 95% CI = 0.43-0.78) than that in the group with low stress levels (IRR = 1.03, 95% CI = 0.83-1.28), especially among men (see Table 4 for estimates).
TABLE 4: Physical activity and IRR of URTI by levels of perceived stress.
Because pollen allergy can mimic a URTI, we excluded the last two follow-ups, which corresponded to pollen season, for those reporting a pollen allergy. The results were similar to the findings from the main analyses: high physical activity in combination with high perceived stress was associated with a decreased risk of URTI among all participants (IRR = 0.61, 95% CI = 0.44-0.82), especially men (IRR = 0.43, 95% CI = 0.26-0.72). There was a weaker association among women (IRR = 0.75, 95% CI = 0.51-1.10) (compare with Table 4). In addition, because physical activity habits and perceived stress might vary with season during our study, we also conducted an analysis in which we included only URTI cases occurring during the first 3 follow-ups. The results were again similar to the main analysis when comparing high physical activity in combination with high stress to low physical activity and high stress among all participants (IRR = 0.50, 95% CI = 0.34-0.74), men (IRR = 0.37, 95% CI = 0.20-0.69), and women (IRR = 0.64, 95% CI = 0.39-1.03) (compare with Table 4). We found no indication of any important effect measure modification on physical activity by age, body mass index, smoking, month, or contact with children.
We investigated the shape of the dose-response relationship between physical activity and stress on incidence of URTI using spline regression as a smoothing method to depict the trend in effect over level of physical activity. Results were similar for both the categorical and the spline analyses. Figure 2 shows smoothed incidence rates of URTI for high and low stress groups over varying levels of MET-hours per day for all participants (A) and for men (B) and women (C), respectively. In the graph including all participants, high levels of physical activity appeared protective for those experiencing either low or high levels of perceived stress, but the protective effect was greater for those experiencing high levels of perceived stress. For low stress, the apparent protective effect of physical activity (shown as a decline in the rate) comes with lower levels of physical activity and then levels out. For high stress, the protective effect of physical activity comes with higher levels of physical activity but then increases more steadily. For men, the protective effect of physical activity appears to be very small, if any, among those experiencing low levels of perceived stress but considerably larger among those experiencing high levels of perceived stress. For women, the difference in stress levels does not appear to influence greatly the effect of physical activity. There is considerable overlap in the 95% CI for high and low stress, so our inference about trends is tentative and limited to a broad characterization.
DISCUSSION
In this population-based Web cohort, we found that high levels of physical activity were associated with a reduced risk of self-reported URTI. The association between physical activity and URTI was seen for both men and women, across all ages, and for both smokers and nonsmokers. We found protective effects of similar magnitude for both vigorous (≥6 METs) and nonvigorous (3-5 METs) physical activity. In addition, we found that participants, especially men, reporting high levels of perceived stress seemed to benefit more from high physical activity than those reporting lower levels of perceived stress.
Our result confirm the findings from other studies, suggesting an inverse association between moderate to high levels of physical activity and URTI (16,20,26,28). However, we did not observe a J-shaped association between physical activity and URTI risk, as suggested by Nieman (24). This discrepancy might be explained by the paucity of participants in our study whose physical activity level corresponded to neither the intensity nor duration to that of a marathon runner, for whom increased risk of URTI has previously been observed (24). A participant classified as having a high physical activity (>55 MET·h·d−1) in our study would for example be someone with a sedentary job that goes jogging or to the gym for an hour each day and is moderately active the rest of the day (for example, takes the stairs, does household chores, or actively plays with the children).
A previous epidemiological study (34) as well as experimental studies (for a review, see 5) have shown a progressively increasing risk of URTI with increasing levels of psychological stress. A cohort study among 1149 university employees reported a 2.5-fold increased risk of common cold among participants reporting high perceived stress levels (34). It has also been shown that a high level of perceived stress is associated with increased symptoms of illness and increased production of interleukin 6 in response to a viral challenge (6). We used the 14-item validated version of the Perceived Stress Scale by Cohen et al. (7) to assess perceived stress during the previous month, but we did not find an overall effect of high perceived stress on URTI, nor did results change when we only included the first three follow-ups to minimize the effect of potential change in perceived stress occurring during the study. However, it is possible that perceived stress assessed in this population-based study was not severe or long term enough to affect immune function. In addition, we did see an interaction between stress and physical activity on the immune system, and this interaction has been previously observed in animal studies, revealing that regular, moderate physical activity reduces the negative effects of an acute stressor on the immune response (for a review, see Greenwood and Fleshner [10]).
We observed a difference between men and women regarding the relation between perceived stress levels and physical activity. This difference might possibly be explained by the proposed difference in physiological and behavioral responses to stress for men and women (35). The "fight-or-flight" response to stress, although present in both men and women, is proposed to be stronger in men and a "tend-and-befriend" response more common in women in response to stress (35). These response differences could explain why men might benefit more from physical activity while under stress than women.
We saw a decline in URTI with increasing age, which is in line with previous studies (23). As we get older, we are more likely to encounter viruses that our immune system recognizes, which may explain part of this decline. Also, there is a documented overall decline in immune function among elderly (age <65 yr) (for a review, see Panda et al. 29), but the participants in our study were most likely too young (age 20-60 yr) for this decline to be observed.
Previous studies have found that physical activity is associated with lower influenza-related mortality (40). However, when we separated URTI into systemic URTI including fever (i.e., sign of influenza) and nonsystemic URTI without fever (i.e., indicates common cold) in an attempt to differentiate between common cold and influenza, we found no protective associations of physical activity on systemic URTI, but these results are based on a limited number of participants. Also, participants in our study were not specifically asked to check their temperature, and therefore we might have overestimated or underestimated the number of cases with fever. Furthermore, the question regarding weakened immune system (because of, e.g., medication or chronic disease) has not been validated. However, only 2% of the participants reported having a weakened immune system, and it is highly unlikely that this variable would have changed the results significantly.
Self-diagnosis of URTI has been shown to be reliable in adult patients (13). However, Spence et al. (33) showed that not all self-reported URTI have a verified pathogenic cause, making it difficult to distinguish between symptoms and infection. Because we studied self-reported URTI symptoms and not verified infections, misclassification of disease is possible, but it is unlikely that this has affected the result greatly after controlling for both pollen allergy and smoking, factors known to be associated with URTI-like symptoms. Furthermore, we assessed and adjusted for potential confounders in the final statistical model, but unmeasured aspects of a healthy lifestyle might cause a bias because it is hard to single out the effect of physical activity.
Physical activity is an inexpensive, modifiable lifestyle factor that has been associated with reduced risks of a large number of diseases, including type 2 diabetes (15), cardiovascular diseases (36), certain cancers (41), and overall mortality (37). Our findings confirm that high physical activity is associated with a reduced risk of URTI and indicate that the effect of physical activity in preventing URTI is greater among highly stressed people, especially for men.
CONCLUSIONS
This population-based study used frequent follow-ups to assess URTI symptoms. It shows that people reporting moderate to high total physical activity had a lower risk of URTI. In addition, highly stressed people, particularly men, might benefit more from physical activity than those with low stress levels.
This work was supported by a grant from the Swedish Council for Working Life and Social Research (grants 2002-04-23, 2003-0786, and 2006-1424); the Swedish Research Council (grant 2005-7102); and the Osher Center for Integrative Medicine, Karolinska Institutet.
The authors thank Anna Johansson for thorough statistical advice and programming assistance.
The results of the present study do not constitute endorsement by the American College of Sports and Medicine.
REFERENCES
1. Aadahl M, Kjaer M, Jorgensen T. Associations between overall physical activity level and cardiovascular risk factors in an adult population.
Eur J Epidemiol. 2007;22(6):369-78.
2. Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities.
Med Sci Sports Exerc. 1993;25(1):71-80.
3. Ainsworth BE, Haskell WL, Whitt MC, et al. Compendium of physical activities: an update of activity codes and MET intensities.
Med Sci Sports Exerc. 2000;32(9 suppl):S498-516.
4. Balter KA, Balter O, Fondell E, Lagerros YT. Web-based and mailed questionnaires: a comparison of response rates and compliance.
Epidemiology. 2005;16(4):577-9.
5. Cohen S. Keynote presentation at the Eight International Congress of Behavioral Medicine: the Pittsburgh common cold studies: psychosocial predictors of susceptibility to respiratory infectious illness.
Int J Behav Med. 2005;12(3):123-31.
6. Cohen S, Doyle WJ, Skoner DP. Psychological stress, cytokine production, and severity of upper respiratory illness.
Psychosom Med. 1999;61(2):175-80.
7. Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress.
J Health Soc Behav. 1983;24(4):385-96.
8. Fleshner M. Physical activity and stress resistance: sympathetic nervous system adaptations prevent stress-induced immunosuppression.
Exerc Sport Sci Rev. 2005;33(3):120-6.
9. Georgiades A, Sherwood A, Gullette EC, et al. Effects of exercise and weight loss on mental stress-induced cardiovascular responses in individuals with high blood pressure.
Hypertension. 2000;36(2):171-6.
10. Greenwood BN, Fleshner M. Exercise, learned helplessness, and the stress-resistant brain.
Neuromolecular Med. 2008;10(2):81-98.
11. Greenwood BN, Foley TE, Day HE, et al. Freewheel running prevents learned helplessness/behavioral depression: role of dorsal raphe serotonergic neurons.
J Neurosci. 2003;23(7):2889-98.
12. Grimsmo A, Hagman E, Faiko E, Matthiessen L, Njalsson T. Patients, diagnoses and processes in general practice in the Nordic countries: an attempt to make data from computerised medical records available for comparable statistics.
Scand J Prim Health Care. 2001;19(2):76-82.
13. Gwaltney JM Jr. The common cold. In: Mandell GL, Douglas RG, Bennett JE, Dolin R, editors.
Principles and Practice of Infectious Diseases. Philadelphia (PA): Churchill Livingstone; 2000. p. 651-6.
14. Hemila H, Virtamo J, Albanes D, Kaprio J. Physical activity and the common cold in men administered vitamin E and beta-carotene.
Med Sci Sports Exerc. 2003;35(11):1815-20.
15. Kelley DE, Goodpaster BH. Effects of physical activity on insulin action and glucose tolerance in obesity.
Med Sci Sports Exerc. 1999;31(11 suppl):S619-23.
16. Kostka T, Drygas W, Jegier A, Praczko K. Physical activity and upper respiratory tract infections.
Int J Sports Med. 2008;29(2):158-62.
17. Lagerros YT, Mucci LA, Bellocco R, Nyren O, Balter O, Balter KA. Validity and reliability of self-reported total energy expenditure using a novel instrument.
Eur J Epidemiol. 2006;21(3):227-36.
18. Linde A, Brytting M, Johansson M, Wiman A, Hogberg L, Ekdahl K.
The National Influenza Reference Centre. Swedish Institute for Infectious Disease Control. Annual Report September 2003-August 2004. 2004. 13 p.
19. Martin SA, Pence BD, Woods JA. Exercise and respiratory tract viral infections.
Exerc Sport Sci Rev. 2009;37(4):157-64.
20. Matthews CE, Ockene IS, Freedson PS, Rosal MC, Merriam PA, Hebert JR. Moderate to vigorous physical activity and risk of upper-respiratory tract infection.
Med Sci Sports Exerc. 2002;34(8):1242-8.
21. Messerer M, Johansson SE, Wolk A. The validity of questionnaire-based micronutrient intake estimates is increased by including dietary supplement use in Swedish men.
J Nutr. 2004;134(7):1800-5.
22. Michels KB, Holmberg L, Bergkvist L, Ljung H, Bruce A, Wolk A. Dietary antioxidant vitamins, retinol, and breast cancer incidence in a cohort of Swedish women.
Int J Cancer. 2001;91(4):563-7.
23. Monto AS, Sullivan KM. Acute respiratory illness in the community: frequency of illness and the agents involved.
Epidemiol Infect. 1993;110(1):145-60.
24. Nieman DC. Exercise, upper respiratory tract infection, and the immune system.
Med Sci Sports Exerc. 1994;26(2):128-39.
25. Nieman DC. Current perspective on exercise immunology.
Curr Sports Med Rep. 2003;2(5):239-42.
26. Nieman DC, Henson DA, Gusewitch G, et al. Physical activity and immune function in elderly women.
Med Sci Sports Exerc. 1993;25(7):823-31.
27. Nieman DC, Nehlsen-Cannarella SL, Markoff PA, et al. The effects of moderate exercise training on natural killer cells and acute upper respiratory tract infections.
Int J Sports Med. 1990;11(6):467-73.
28. Nieman DC, Pedersen BK. Exercise and immune function: recent developments.
Sports Med. 1999;27(2):73-80.
29. Panda A, Arjona A, Sapey E, et al. Human innate immunosenescence: causes and consequences for immunity in old age.
Trends Immunol. 2009;30(7):325-33.
30. Rothman KJ, Greenland S, Lash TL, editors. Introduction to stratified analysis. In:
Modern Epidemiology. Philadelphia (PA): Lippincott, Williams & Wilkins; 2008. p. 258-82.
31. Royston P. Multivariable modeling with cubic regression splines: a principled approach.
Stata J. 2007;7(1):45-70.
32. Segerstrom SC, Miller GE. Psychological stress and the human immune system: a meta-analytic study of 30 years of inquiry.
Psychol Bull. 2004;130(4):601-30.
33. Spence L, Brown WJ, Pyne DB, et al. Incidence, etiology, and symptomatology of upper respiratory illness in elite athletes.
Med Sci Sports Exerc. 2007;39(4):577-86.
34. Takkouche B, Regueira C, Gestal-Otero JJ. A cohort study of stress and the common cold.
Epidemiology. 2001;12(3):345-9.
35. Taylor SE, Klein LC, Lewis BP, Gruenewald TL, Gurung RA, Updegraff JA. Biobehavioral responses to stress in females: tend-and-befriend, not fight-or-flight.
Psychol Rev. 2000;107(3):411-29.
36. Thompson PD, Buchner D, Pina IL, et al. Exercise and physical activity in the prevention and treatment of atherosclerotic cardiovascular disease: a statement from the Council on Clinical Cardiology (Subcommittee on Exercise, Rehabilitation, and Prevention) and the Council on Nutrition, Physical Activity, and Metabolism (Subcommittee on Physical Activity).
Circulation. 2003;107(24):3109-16.
37. Trolle-Lagerros Y, Mucci LA, Kumle M, et al. Physical activity as a determinant of mortality in women.
Epidemiology. 2005;16(6):780-5.
38. Twisk JWR, editor. In:
Applied Longitudinal Data Analysis for Epidemiology: A Practical Guide. Cambridge (UK): Cambridge University Press; 2003. 301 p.
39. Willett WC, Stampfer MJ. Implications of total energy intake for epidemiologic analyses. In: Willett WC, editor.
Nutritional Epidemiology. New York (NY): Oxford University Press; 1998. p. 273-301.
40. Wong CM, Lai HK, Ou CQ, et al. Is exercise protective against influenza-associated mortality?
PLoS One. 2008;3(5):e2108.
41. World Cancer Research Fund., American Institute for Cancer Research, editor. In:
Food, Nutrition, Physical Activity, and the Prevention of Cancer: A Global Perspective. Washington, DC: WCRF/AICR; 2007. 517 p.
