Following Cancer Prevention Guidelines Reduces Risk of Cancer, Cardiovascular Disease, and All-Cause Mortality

This article is featured in Highlights of This Issue, p. 1057

Many health organizations publish guidelines on nutrition, physical activity, and other modifiable behaviors known to affect risk of cancer, cardiovascular disease (CVD), and other major illnesses (1–6). These guidelines seek to summarize and communicate current scientific knowledge about behaviors that can reduce the risk of specific diseases and promote overall health. For example, the American Cancer Society (ACS) guidelines on nutrition and physical activity (1) are targeted specifically for cancer prevention, although they share many features with other organizations' guidelines (3–5).

Few studies have evaluated the relationship between adherence to multiple health recommendations and risk of developing (7) or dying from major chronic diseases (8–12), and only 1 specifically tested cancer prevention recommendations (12). Most studies combined tobacco cessation with other modifiable risk factors and reported all-cause death or disease incidence rates that were 21% to 78% lower among people who reported several healthful behaviors compared with none (7–11). Given the overwhelmingly large detrimental effects of tobacco use, and the fact that 80% of Americans have either stopped smoking or never started (13), the potential impact of recommended behaviors, separate from smoking, on improving health outcomes is of considerable interest.

This study examined whether adherence to the ACS guidelines on body weight, physical activity, diet, and alcohol consumption was associated with risk of death from all causes, CVD, and cancer in never and former smokers in the large Cancer Prevention Study II (CPS-II) Nutrition Cohort (14).

The CPS-II Nutrition Cohort is a prospective study of cancer incidence and mortality among 86,403 men and 97,786 women (14). At enrollment in 1992 or 1993, participants were ages 50 to 74 years and completed a 10-page confidential self-administered questionnaire that included questions on demographic, medical, lifestyle, and dietary factors. Nutrition cohort participants are a subgroup of the approximately 1.2 million members of the CPS-II cohort, a prospective study of cancer mortality established by the ACS in 1982 (15), who resided in 21 states with population-based cancer registries (14). The Institutional Review Board of Emory University School of Medicine, Atlanta, GA, approves all aspects of the CPS-II Nutrition Cohort.

The vital status of participants was determined through December 31, 2006 by linkage with the National Death Index (16). The underlying causes of death were coded according to the International Classification of Diseases, 9th and 10th revisions (ICD-9 and ICD-10; refs. 17–18). Cancer deaths included ICD-9 codes 140–195 and 199–208, or ICD-10 codes C00-C76 and C80-C97. CVD deaths were defined as ICD-9 codes 390–459, or ICD-10 codes I00-I99.

To quantify lifestyle behaviors consistent with recommendations from the 2006 ACS guidelines on nutrition and physical activity for cancer prevention (1), an a priori score was developed where each of 4 major guidelines was weighted equally from 0 to 2, where 0 represented not meeting the recommendation, 1 represented partial compliance, and 2 represented full compliance. This scoring approach allowed for a more complete examination of adherence than yes/no scoring. The overall adherence score was the sum of the 4 components and ranged from 0 for those who did not adhere to any of the guidelines to 8 for those who were fully compliant.

The first recommendation, to “maintain a healthy weight throughout life”, was scored by height and weight reported in 1982 (from the parent CPS-II cohort) and weight in 1992 to take advantage of the fact that information on adult body weight was reported 10 years prior to baseline, thus providing a long-term measure of body mass index (BMI, in kg/m²) in adulthood. The optimal score (2) was given to participants whose BMI was within the normal range (18.5–<25 kg/m²) at both time points. A score of zero was given to individuals who were obese (>30 kg/m²) at both time points, or obese at 1 time point and overweight (25–<30 kg/m²) at the other time point. A score of 1 was given to all others. Underweight individuals (<18.5 kg/m²) were excluded.

The second recommendation is to “adopt a physically active lifestyle with 30 minutes or more of moderate to vigorous intentional physical activity at least 5 days/week; 45 to 60 minutes are preferable” (1). Participants reported the average time spent, per week, in 7 recreational activities at baseline in 1992. These were converted to metabolic equivalent (MET) scores (19). The minimum standard of 30 minutes on 5 days (2.5 hours/week) of moderate activity (3.5 METs) is equal to 8.75 MET-hours/week. One hour per day, 5 days a week (5.0 hours/week), of moderate activity (3.5 METs) is equal to 17.5 MET-hours/week. Therefore, MET-hours/week less than 8.75 received a score of 0, 8.75 to <17.5 earned a score of 1, and ≥17.5 MET-hours earned a score of 2 for reaching “preferable” levels.

The 3rd recommendation is to “consume a healthy diet with an emphasis on plant sources” with specific recommendations as described next (1). Diet was assessed at baseline by a validated 68-item food frequency questionnaire (FFQ; refs. 14, 20, 21). To capture the totality of the recommended diet pattern, sex-specific diet quality scores were first created, which ranged from 0 to 9 (0–3 for each of 3 diet recommendations). For the recommendation “eat 5 or more servings of a variety of vegetables and fruits each day,” participants received a point if they consumed ≥5 servings/day and an additional 1 or 2 points for being in the 2nd or 3rd tertile of unique fruits and vegetables consumed per month (a measure of “variety”). For the recommendation to “choose whole grains in preference to processed (refined) grains,” the percentage of grains consumed as whole grains (daily servings of whole grains/daily servings of whole grains plus refined grains) was assigned a score of 0 to 3 for the quartile distribution (highest quartile = 3). The recommendation to “limit consumption of processed and red meats” was assigned a score of 0 to 3 for the quartile distribution of servings consumed per week (lowest quartile = 3). All diet recommendations were summed and then collapsed to 0 to 2 points: those who scored 0 to 2 on the 9-point diet score were given a score of 0, 3 to 6 were scored as 1, and 7 to 9 were scored as 2.

The ACS recommendation on alcohol states: “if you drink, limit consumption to 1 drink/day for women and 2 drinks/day for men” (1). Because the relationship with alcohol consumption is different for CVD than for cancer (22, 23), 2 approaches were used to score compliance with the alcohol recommendation for cancer. For all outcomes, the lowest score (0) was given to men who consume more than 2 drinks/day and women who consume more than 1 drink per day, nondrinkers were assigned a score of 1, and those who drink only moderately (>0 but ≤1 drink/day for women and ≤2 drinks/day for men) received a score of 2. In the second analysis of cancer mortality, the alcohol score was changed such that moderate drinkers were assigned the score of 1 and nondrinkers were assigned the score of 2.

In the analysis of overall scores, those in the tails of the distribution (0, 1, and 2 for lowest scores; 7 and 8 for highest) were combined because of small numbers.

Men and women were excluded from this analysis if at baseline, they reported a history of cancer (except nonmelanoma skin cancer; n = 23,630), myocardial infarction (n = 11,431), or stroke (n = 2,483); an extreme value for height or weight in 1982 or 1992/1993 (top and bottom 0.1%; n = 4,295), and below-normal weight (BMI < 18.5 kg/m²) in 1982 or 1992/1993 (n = 2,510; ref. 6). Current smokers (n = 11,552) and individuals with missing information on smoking (n = 1,297), physical activity (n = 2,048), and alcohol use (n = 5,394) or who poorly completed the FFQ (n = 7,583) were excluded. The final analytic cohort included 50,727 men and 61,239 women. After 14 years of follow-up, the total number of deaths was 16,982 (10,369 men, 6,613 women), including 5,628 deaths due to CVD (3,666 men, 1,962 women) and 5,874 deaths due to cancer (3,318 men, 2,556 women).

Participants contributed follow-up time from the date of return of the baseline questionnaire in 1992 or 1993 to the date of death or December 31, 2006. Cox proportional hazards regression analysis (24) was used to estimate the relative risk (RR) of death and 95% CI by using the PHREG procedure in SAS, version 9.1 (SAS Institute).

Age adjustment was achieved by stratifying on single year of age within each Cox model. Covariates included education (less than high school graduate, (ref.) high school graduate, some college/trade, and college graduate) and smoking history (never and former (classified by years since quitting ≤5, >5–10, >10–15, >15–20, >20–25, >25) (ref.), and unknown years since quitting). Other covariates considered (nonsteroidal anti-inflammatory drug use, multivitamin use, race, menopausal hormone therapy, and energy) did not change the associations. Although tests of the proportional hazards assumption suggested that the relationship of the guidelines score with all-cause and CVD mortality varied quantitatively over time (P < 0.05), the associations were qualitatively similar; therefore, we present results including the full follow-up period.

Associations were examined separately in never and former smokers (current smokers were excluded). In addition, associations between the non-BMI–related guidelines (i.e., physical activity, diet, and alcohol consumption) and mortality were examined in individuals whose BMI was consistently heavy, normal, or other. For this analysis, a modified 0 to 6 point score, excluding BMI, was used. The likelihood ratio test was used to test for multiplicative interactions comparing Cox multivariable models with and without cross-product terms where the score was modeled as a continuous variable. All tests of statistical significance were 2-sided.

The mean (SD) age of men and women at baseline was 63.6 (6.0) and 61.9 (6.5) years. Only 3.5% of men and 4.0% of women met the optimal recommendation for each component of the guidelines and were assigned a score of 8. Table 1 shows age-standardized demographic characteristics in relation to the guidelines score in men and women. Those with higher scores were slightly older and more educated. Men with higher scores were more likely to be never smokers, whereas women with higher scores were more likely to be former smokers. The average grams/day of ethanol consumed was at least 3 times higher among drinkers with low versus high scores. The prevalence of diabetes and hypertension was lower among participants with higher scores.

Table 2 illustrates the all-cause and cause-specific mortality risk associated with the guidelines score. Higher scores were strongly inversely associated with lower total mortality and cause-specific mortality in both sexes. Associations for CVD were strongly linear, whereas those for cancer seemed to level off after a score of 4 in men and 5 in women. Results were similar for cancer mortality when nondrinkers received a score of 2 for the alcohol component.

The associations of individual components of the score (mutually adjusted) are provided in Table 3 (men) and Table 4 (women). As compared with participants with consistently high BMI, men and women with a consistently normal BMI were at lowest risk of death, especially from CVD; the “other BMI combinations” group (score = 1) was also at lower risk. Meeting the physical activity guidelines was inversely related to all-cause and CVD mortality, but was not independently related to cancer mortality. Moderate alcohol consumption, compared with consumption above recommended limits, was inversely associated with risk of all outcomes in men, although the confidence limits included 1.0 for CVD mortality. Men who reported no alcohol consumption had a lower risk of dying from cancer than heavier drinkers. In contrast, nondrinking was associated with higher risk of all-cause and CVD mortality, especially in women. The recommended diet pattern was associated with a significantly lower risk for all outcomes, except for CVD in men (P_trend = 0.2).

As shown in Figure 1, compared with a referent group of never smokers with a low guidelines score, both never and former smokers who reported greater adherence to the guidelines had a lower risk of all-cause, CVD, and cancer mortality. For cancer mortality, although the trend was stronger in never than former smoking women, and in former than never smoking men, associations stratified by smoking status were not statistically significant (P_interaction > 0.1).

We examined the role of lifestyle behaviors apart from BMI (diet, physical activity, and alcohol) by using 2 approaches. When BMI was removed from the score (new range = 0–6) and instead included as a covariate (BMI in 1982, <25, 25–<30, and >30 kg/m²), RRs (95% CI) for scores of 5 or 6 vs. 0 or 1 in women were 0.73 (0.66–0.80) for total mortality (P_trend < 0.0001), 0.60 (0.51–0.71) for CVD mortality (P_trend <0.0001), and 0.86 (0.75–1.00) for cancer mortality (P_trend = 0.02). For men, estimates were 0.70 (0.65–0.75) for total mortality (P_trend < 0.0001), 0.72 (0.63–0.81) for CVD mortality (P_trend <0.0001), and 0.80 (0.70–0.90) for cancer mortality (P_trend<0.0001). In addition, adherence to the diet, physical activity, and alcohol guidelines was associated with reduced risk within each BMI score category (P_interaction > 0.10 for all endpoints, data not shown).

Sensitivity analyses found similar results after excluding participants with previously diagnosed diabetes (data not shown).

After tobacco avoidance and cessation, weight control, physical activity, dietary choices, and moderation of alcohol intake are the primary lifestyle strategies to prevent cancer (1–2). However, few Americans realize the potential of these other lifestyle modifications for cancer prevention (25). Guidelines for cancer prevention largely overlap with those for prevention of diabetes (4) and CVD (3, 26), but some differences, for example, in dietary recommendations, exist. In this large cohort study of never and former smokers, men and women whose behaviors were most consistent with cancer prevention guidelines (1) had a 42% lower risk of death from any cause during the 14-year follow-up period than individuals whose behaviors were least consistent. The risk of death from cancer or CVD was also significantly lower among those reporting behaviors consistent with the guidelines. Maintenance of a healthy BMI seemed to be the component most strongly associated with lower mortality, but all lifestyle behaviors significantly contributed to mortality risk reduction.

The most important strategy to reduce the risk of cancer (27) and many other chronic diseases is tobacco avoidance. Most other studies of all-cause mortality or chronic disease incidence in relation to lifestyle behaviors included tobacco smoking in the risk score (7–11). In these studies, tobacco was often the strongest predictor of risk (8, 9); however, other lifestyle behaviors also reduced risk in never smoking women (10, 12) or men and women combined (7). In the current study, the largest of its type, and the first to examine associations separately in never and former smoking men and women, we confirmed that following healthy lifestyle recommendations for cancer prevention reduce the risk of premature death in these groups. For cancer mortality, inverse associations, albeit not consistently dose related, were observed in all groups.

On the basis of extensive evidence relating excess adiposity to increased incidence and death rates from cancer (28, 29) and CVD (30, 31), the ACS highlights maintenance of a healthy body weight as the first in its list of guidelines on nutrition and physical activity (1). In the current study, maintenance of a healthy body weight for 10 years prior to and including baseline was strongly associated with lower mortality risk than individuals who were consistently heavy. The importance of this issue cannot be overstated as 68% of American adults are overweight or obese (32). In addition, this study found that the inverse associations with healthy patterns of diet, physical activity, and alcohol consumption were observed among those who were consistently lean or consistently heavy. Individuals who are normal weight or overweight might also have metabolic (33) or other risk factors for cancer that may be modified by these behaviors. Ultimately, the current study highlights the benefits of following these recommendations among all individuals.

In this study, adherence to the multiple dietary recommendations was assessed by a single dietary pattern variable. This approach aggregates the influence of individual dietary factors and captures additive and interactive effects of nutritional constituents (34). A diet low in red and processed meat, with a greater proportion of whole versus refined grains, and high in a variety of vegetables and fruits, was independently associated with lower mortality from all-cause, cancer, and CVD mortality than participants with less optimal diet patterns; however, the association was not statistically significant for CVD in men. Results of other studies examining various diet patterns and cancer outcomes have been inconsistent (8–11, 35–40), possibly due to “dilution” of associations by the inclusion of dietary components (e.g., trans fats), as yet not shown to be cancer-related, in the scores. Few other studies specifically tested cancer prevention guidelines (12). Cerhan and colleagues (12) reported a 35% to 43% higher risk of total cancer incidence and death among postmenopausal women with the worst (0, 1) versus the best (6–9) scores on the basis of diet, body weight, and physical activity recommendations from the 1997 World Cancer Research Fund/American Institute for Cancer Research (41).

Physical activity was not independently associated with cancer mortality in our study, but was associated with lower CVD and all-cause mortality. Most physical activity in this older largely retired population involved walking rather than more vigorous activities. Even though relatively low levels of physical activity are known to be associated with lower risk of colon cancer in this (42) and other (29) cohorts, higher levels of physical activity may be necessary for overall cancer prevention (2).

Heavy alcohol consumption is associated with an elevated risk of several cancers (2, 43), injuries, cirrhosis, and possibly hemorrhagic stroke, but moderate consumption is associated with lower risk of coronary heart disease (23). The scoring of “optimal” alcohol consumption in other studies of healthy lifestyle scores and total and cause-specific mortality has varied. For example, 2 studies considered moderate consumption to be optimal (versus none or excessive consumption; refs. 8, 10), 2 other studies considered moderate or none to be optimal (versus excessive; refs. 11–12), whereas a 5th considered any consumption as optimal (versus none; ref. 9). Other studies did not include alcohol in the score (7, 44). In the current analysis, the benefits of moderate drinking (versus heavy or nondrinking) for CVD and total mortality were clear. For cancer mortality, benefits of either nondrinking or moderate alcohol consumption, compared with heavy drinking, were observed in men. Overall, these findings support cancer prevention recommendations to limit alcohol consumption among those who drink.

Strengths of this study include its large size, prospective design, and detailed assessment of lifestyle factors and smoking history. A limitation is that BMI may be a less useful indicator of adiposity among older individuals who redistribute excess weight centrally—which is also a risk factor for several chronic diseases, independent of BMI (45). The dietary assessment instrument included only 68 line items, but was validated for the food groups used in this analysis (21). Nevertheless, the range of unique fruits and vegetables regularly consumed (0–15) might have underestimated the range of “variety.” Furthermore, assessment of physical activity and diet are subject to considerable measurement error (46), so our risk estimates likely underestimate these associations. Participants in the CPS-II Nutrition Cohort are a relatively health-conscious sample of primarily older Caucasian U.S. adults, and the impact of following these recommendations at younger ages or in other racial/ethnic groups is not known.

In conclusion, these findings provide evidence that adhering to cancer prevention guidelines for body weight, diet, physical activity, and alcohol intake, individually and in aggregate, improve survival among older nonsmoking adults.

No potential conflicts of interest were disclosed.

We thank R. Diver, MPH, and R. Shah, MPH, for additional statistical support.