Preventable incidence and mortality of carcinoma associated with lifestyle factors among whites in the United States

Abstract

Importance

Lifestyle factors are important for cancer development. However, a recent study has been interpreted to suggest that random mutations during stem cell divisions are the major contributor to human cancer.

Objective

To estimate the proportion of cases and deaths of carcinoma (all cancers except skin, brain, lymphatic, hematologic, and non-fatal prostate malignancies) among whites in the United States that can be potentially prevented by lifestyle modification

Design

Prospective cohort study

Setting

Nurses’ Health Study, Health Professionals Follow-up Study, and national cancer statistics

Participants

16,531 women and 11,731 men who had a healthy lifestyle pattern (low-risk group), and the remaining 73,040 women and 34,608 men (high-risk group).

Exposure

The healthy lifestyle pattern was defined as never or past smoking with pack-years less than 5, no or moderate alcohol drinking (≤1 drink/day for women, ≤2 drinks/day for men), body mass index ≥18.5 and <27.5 kg/m², and weekly aerobic physical activity of at least 75 vigorous-intensity or 150 moderate-intensity minutes.

Main outcomes and measures

We calculated the population attributable risk (PAR) by comparing incidence and mortality of total and major individual carcinomas between the low- and high-risk groups. We further assessed the PAR at the national scale by comparing the low-risk group to the US population.

Results

Within our cohorts, the PARs for incidence and mortality of total carcinoma were 25% and 48% in women, and 33% and 44% in men, respectively. For individual cancers, the PARs in women and men were 82% and 78% for lung, 29% and 20% for colon and rectum, 30% and 29% for pancreas, and 36% and 44% for bladder. Similar estimates were obtained for mortality. The PARs were 4% and 12% for breast cancer incidence and mortality, and 21% for fatal prostate cancer. Substantially higher PARs were obtained when comparing our low-risk group to the general US population. For example, the PARs in women and men were 41% and 63% for incidence of total carcinoma, and 60% and 59% for colorectal cancer, respectively.

Conclusions and relevance

A substantial cancer burden may be prevented through lifestyle modification. Primary prevention should remain a priority for cancer control.

Introduction

Cancer is the second leading cause of death in the United States, with 1.6 million new cancer cases and a half million cancer deaths projected to occur in 2015.¹ The age-standardized mortality rate from cancer has decreased from 199 to 163 (per 100 000) between 1969 and 2013 according to the 2000 US standard population.² However, this decline (17.9%) has been modest compared to the dramatic decrease in heart disease mortality (67.5%) during the same period, highlighting the need for further efforts in cancer prevention and treatment.

Epidemiologic studies have established several lifestyle factors that increase cancer risk, such as smoking, alcohol use, obesity, and physical inactivity.³ However, this substantial body of knowledge has been challenged by a recent study,⁴ which found a high correlation between the number of stem cell divisions of a given tissue and the lifetime risk of cancer in that tissue. This finding led some to conclude that only a third of the variation in cancer risk among tissues is attributable to environmental factors or inherited predispositions, while the majority is due to random mutations arising during stem cell divisions, so-called “bad luck”. This study has been widely covered by the press and created confusion for the public regarding the preventability of cancer. Many arguments against the “bad luck” hypothesis have been made,^5–12 including the notion that external environmental factors may influence cancer development through promotion of DNA damage;¹³ yet none of these reports has provided original data to assess the preventability of cancer through modification of extrinsic factors.

Therefore, we estimated the contributions of common lifestyle factors to cancer burden by comparing cancer incidence and mortality between the participants who had a healthy lifestyle (low-risk group) to those who did not (high-risk group) in the two nationwide cohorts. We further explored the potential capability of lifestyle modification for cancer prevention at the national scale, by comparing the low-risk subgroup of our cohorts to the US population. Because our cohorts’ participants were predominantly whites, to avoid any influence of different racial distributions on the comparison to the general population, we only included whites in the analysis.

Methods

Study population

The Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (HPFS) and are two ongoing US cohorts that enrolled 121,700 registered female nurses aged 30–55 years with about 70% response rate in 1976 and 51,529 male health professionals aged 40–75 years with about 25% response rate in 1986, respectively. Similar follow-up procedures have been used in the two cohorts.^14,15 In brief, participants completed a detailed questionnaire about their medical history and lifestyle at baseline, and every two years thereafter. Dietary intake was assessed using validated food frequency questionnaires (FFQs) every four years. The response rates have been 95.4% in the NHS and 95.9% in the HPFS for each of the questionnaires though 2010. In the present study, we used 1980 for the NHS and 1986 for the HPFS as baseline, when we first collected detailed lifestyle data.

We identified 16,531 women from the NHS and 11,731 men from the HPFS as the low-risk group who met the following four lifestyle criteria: never smoking or past smoking with pack-years of smoking less than 5 years, no or moderate alcohol drinking (≤1 drink/day for women, ≤2 drinks/day for men) as recommended by the Dietary Guidelines for Americans,¹⁶ BMI ≥18.5 and <27.5 kg/m², and weekly aerobic physical activity of at least 75 vigorous-intensity or 150 moderate-intensity minutes (7.5 metabolic-equivalent [METs] hours per week) as recommended by the 2008 Physical Activity Guidelines for Americans.^17,18 The remaining 73,040 women and 34,608 men who did not meet all the four criteria and had complete lifestyle data were classified into the high-risk group (see flowchart in eFigure 1). This study was approved by the Institutional Review Board at Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health. Written informed consent was obtained from all study participants.

Lifestyle assessment

Height, body weight, smoking, and physical activity were self-reported on biennial questionnaires. Physical activity was calculated by summing the products of time spent on a variety of leisure-time activities with the average MET for that activity.¹⁹ Alcohol use was self-reported every four years on the FFQs.²⁰ We calculated the overall dietary score according to the Alternate Healthy Eating Index (AHEI), which is designed to target food choices and macronutrient sources associated with reduced chronic disease risk.²¹ Detailed sources of lifestyle data in the US are provided in the eMethods.

Outcome ascertainment

The primary outcomes of this study were incidence and mortality of total and major individual carcinomas. For total carcinoma, we excluded from all cancers those in the skin, brain, lymphatic, and hematopoietic tissues, because these cancers likely have other strong environmental causes than the ones considered in the current study, such as UV exposure, infections, radiation, and exposures to carcinogenic substances. The total carcinomas we studied account for about 90% of all cancer deaths in the US whites. For individual carcinomas, we included those with at least ten cases occurring in our low-risk subpopulation. Given the concern about overdiagnosis for indolent prostate cancer by prostate-specific antigen (PSA) screening,²² we only included fatal prostate cancer in our analysis. The number of cases for each cancer is provided in eTable 1.

In both cohorts, self-reported diagnoses of cancer were obtained on biennial questionnaires, and participants who reported a cancer diagnosis were asked for permission to acquire their medical records and pathologic reports. Study physicians, blinded to exposure information, reviewed medical records to confirm cancer diagnosis. Most of the deaths were identified through family members or the postal system in response to the follow-up questionnaires. We also searched the names of persistent nonresponders in the National Death Index. More than 96% of deaths have been identified using these methods.²³ The cause of death was assigned by study physicians blinded to exposure data.

National incidence and mortality data

We obtained the US cancer incidence²⁴ and mortality²⁵ data from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program. To parallel the age distribution and follow-up time of our cohorts, we selected data collected from participants aged 40 or older, and from 1976 through 2012 for women and from 1986 through 2012 for men. For cancer mortality, the follow-up data were available up to 2011. We only included whites, as we did in our cohort population. All incidence and death rates were age-standardized to the 2000 US standard population using the NCI’s SEER*Stat software (version 8.1.5).²⁶

Statistical analysis

In our cohorts, we calculated person-years of follow-up for each participant from the age at the date of returning the baseline questionnaire until the age at the date of death, loss to follow-up, or end of follow-up (June 1, 2012 for the NHS, January 31, 2012 for the HPFS), whichever came first. For cancer incidence analysis, follow-up was also censored when a participant was diagnosed with any of the cancers under study. Age- and sex-specific rates were calculated for each of the ten age groups (<45, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, 75–79, 80–84, and ≥85 years), and then standardized to the 2000 US standard population.

Our primary outcome measure was the population attributable risk (PAR, %), which included two sets: one was based upon comparison within our cohorts and calculated as the difference in the cancer rates between the low- and high-risk groups divided by the rate in the high-risk group; and the other one was estimated by comparing the cancer incidence and mortality in our low-risk group to the national SEER rates. The PAR can be interpreted as the proportion of cases that would not occur if all individuals adopted the lifestyle of the low-risk population. Details about derivation of the 95% confidence interval (CI) for PAR are provided in the eMethods.

Results

Table 1 shows the comparison of major lifestyle factors in low- and high-risk groups of our cohorts and white Americans. Although diet was not a specific criterion, our low-risk group had a higher AHEI score. The lifestyle profile in the US population was generally even worse than that in our high-risk group.

Table 1.

Comparison of lifestyle factors in the low- and high-risk groups and in the US population in 2010^*

Variable	Women			Men
	Low-risk group	High-risk group	US population	Low-risk group	High-risk group	US population
Smoking†
Never, %	78	31	66	83	26	54
Past, %	22	61	18	17	69	25
Current, %	0	8	16	0	5	21
Met 2008 federal physical activity guideline, %†‡	100	63	47	100	83	54
Body mass index, kg/m2§	23.4	26.7	26.9	24.4	27.3	26.4
Alcohol, drink/day‖	0.1	0.1	0.5	0.4	0.7	1.4
AHEI score§	48.1	46.4	39.0	43.8	40.5	35.7

Abbreviations: AHEI, Alternative Healthy Eating Index.

^*All variables are age-adjusted based on the age distribution of the U.S. population in 2000.

^†Smoking and physical activity data in the US were derived from the National Health Interview Survey, 2012. Past smokers were persons who have smoked at least 100 cigarettes in their lifetime but currently do not smoke at all.

^‡The 2008 federal guidelines recommend that for substantial health benefits, adults should perform at least 150 minutes (2 hours and 30 minutes) a week of moderate-intensity or 75 minutes (1 hour and 15 minutes) a week of vigorous-intensity aerobic physical activity, or an equivalent combination.

^§Mean values were shown.

^‖Median alcohol consumption was shown.

Table 2 shows the age-standardized incidence and mortality rates and the corresponding PAR estimates for total carcinomas among the two risk groups of our cohorts and the whites in the US. The incidence rates of total carcinoma (per 100 000) in the low- versus high-risk groups were 463 versus 618 in women, and 283 versus 425 in men, giving rise to a PAR (95% CI) of 25% (21–29%) in women and 33% (28–38%) in men. A higher PAR was observed for mortality (48% [44–53%] in women and 44% [39–48%] in men). When further comparing our low-risk group to the general US population, we obtained a substantially higher PAR: 41% in women and 63% in men for incidence, and 59% in women and 67% in men for mortality, respectively.

Table 2.

Incidence and mortality rates of total carcinomas in the low- and high-risk groups and in the general US population, and the corresponding estimates of population attributable risk^*

	Rate in the low-risk group (per 100 000)	Rate in the high-risk group (per 100 000)	Rate in the US population (per 100 000)†	PAR (Low-risk vs. high-risk groups) (95% CI)‡	PAR (Low-risk vs. national population) (95% CI)§
Women
Incidence	463	618	789	25% (21–29%)	41% (39–44%)
Mortality	132	256	320	48% (44–53%)	59% (55–62%)
Men
Incidence	283	425	759	33% (28–38%)	63% (60–65%)
Mortality	156	277	470	44% (39–48%)	67% (64–69%)

Abbreviations: CI, confidence interval; PAR, population attributable risk.

^*Total carcinomas include all cancers other than skin cancer, brain cancer, and neoplasms of lymphatic and haematopoietic tissues. Non-fatal prostate cancer is excluded in men. All rates are standardized based on the age distribution of the U.S. population in 2000.

^†US rate was obtained from the Surveillance, Epidemiology, and End Results (SEER) program.

^‡Calculated as the difference in the cancer rates between the low- and high-risk groups divided by the rate of the high-risk group.

^§Calculated as the difference in the cancer rates between the low-risk group and the national rate divided by the national rate.

We then assessed individual cancers (Figures 1 and 2). Within our cohorts, the PARs for incident cancers in women and men were 82% and 78% for lung, 29% and 20% for colon and rectum, 30% and 29% for pancreas, 36% and 44% for bladder, 36% and 4% for kidney, 16% and 38% for oral cavity and pharynx, 27% and 32% for liver, and 62% and 66% for esophagus. The PARs were 4% for breast cancer, and 21% for endometrial, ovarian and fatal prostate cancer. A generally similar PAR was observed for cancer deaths, except a dramatic increase for some sites, including breast (12%), endometrium (49%), kidney in men (48%), and oral cavity and pharynx (75% in women and 57% in men).

Figure 1. Cancer incidence and the corresponding PAR estimates in women (A) and men (B).

The population attributable risks (PARs) comparing the low-risk group to the US population, and comparing the low-risk to the high-risk groups are shown in the plot area as percentage, respectively.

Figure 2. Cancer mortality and the corresponding PAR estimates in women (A) and men (B).

The population attributable risks (PARs) comparing the low-risk group to the US population, and comparing the low-risk to the high-risk groups are shown in the plot area as percentage, respectively.

When the low-risk group was compared to the US population, the resultant PARs were further increased. For example, the PAR rose to 15% for breast cancer incidence and 45% for mortality. For colorectal cancer, the PARs increased to 50–60% in both genders. More detailed data are provided in eTable 2.

Discussion

In the two cohort studies of US whites, we found that overall 20–40% of carcinoma cases and about half of carcinoma deaths can be potentially prevented through lifestyle modification. Not surprisingly, these figures increased to 40–70% when assessed with regard to the general US population of whites, which has a much worse lifestyle pattern than our cohorts. Notably, approximately 80–90% of lung cancer deaths can be avoided if Americans adopted the lifestyle of the low-risk group, mainly by quitting smoking. For other cancers, a range of 10–70% of deaths can be prevented. These results provide strong support for the importance of environmental factors in cancer risk and reinforce the enormous potential of primary prevention for cancer control.

The four factors considered in the current study are among the most prevalent lifestyle factors convincingly linked to various cancers. Smoking contributed to 48.5% of deaths from the 12 smoking-related cancers in the US.²⁷ Heavy alcohol consumption has been causally related to increased risk of cancers in several sites, including colorectum, breast, oral cavity, pharynx, larynx, esophagus, and liver; and possibly to a higher risk of cancers of the lung, pancreas, stomach, and gallbladder.²⁸ Obesity increases risk of cancers in the esophagus (adenocarcinoma), colorectum, pancreas, breast (after menopause), endometrium, kidney, and liver; and probably increases risk of cancers in the ovaries, prostate (advanced only), and gallbladder.³ In contrast, physical activity has been linked to lower risk of cancers in the colorectum, breast, and endometrium.³

These compelling data together with the current study provide strong support that a large proportion of cancers are due to environmental factors and can be prevented by lifestyle modification.¹² Although the stochastic effects of DNA replication error may contribute to the variation in cancer incidence across different tissues,⁴ these influences would be unlikely to explain the wide variation in cancer rates within tissues that have similar lifetime number of stem cell divisions or between individuals with different exposure profiles, or the rapidly increasing burden of cancer in low- and middle-income countries accompanying the global shifts in lifestyle and environmental exposures.^8,29,30

Several previous studies have attempted to quantify the contribution of environmental factors to cancer risk, with the estimated PAR ranging from 30% to 50%.^31–33 A classic approach was often undertaken using the population prevalence of exposure and the relative risk estimate derived from the literature for each risk factor.³⁴ Therefore, several factors can contribute to the variation in PAR estimates, including differences in environmental factors considered in each study, varied definitions and prevalence of exposures, and different sources used to derive the relative risk estimates.

In contrast, our study used a more straightforward approach on the basis of direct comparison of cancer rates between participants with distinct lifestyle profiles. This approach counts on the availability of detailed lifestyle and cancer follow-up data in our two nationwide cohort studies, and circumvents the need for derivation of relative risk estimates for each individual risk factor associated with each cancer. It also takes account of the joint contributions of multiple risk factors that are often difficult, if not impossible, to estimate by the classic approach due to uncertainties about the interactions and complex relationships among risk factors.³⁵ This approach was similar to Doll and Peto’s seminal analysis, in which they compared death rates of the US with the lowest reliably observed rates in other populations and estimated that about 75% of US cancer deaths could be attributed to lifestyle and other environmental factors.³⁶

Our results are also consistent with a recent study showing that cell division-related intrinsic risk factors alone in the absence of environmental risk factors do not confer substantial cancer risk.¹² Through mathematical modeling, that study demonstrates that accumulation of endogenous stem-cell mutation errors, estimated by assuming widely different mutation rates, is not sufficient to account for the observed cancer risk.¹² Taken together with our current empirical data, the two studies provide complementary evidence for the predominant role of extrinsic environmental factors in determining cancer risk.

However, it should be noted that we selected four major cancer risk factors for characterization of the low-risk group to show the preventable potential of cancer rather than to conclude causally that these were the only factors relevant to cancer risk. Therefore, the PARs we calculated might include the contribution by other behaviors that are closely related to the four lifestyle factors considered here and are also important determinants for cancer risk (e.g., diet).

Moreover, we employed a less stringent threshold for characterization of the low-risk profile to allow for a meaningful analysis for some less common cancers. For example, the upper limit for BMI was set at 27.5 rather than 25 kg/m² as commonly used to define normal body weight. Thus, the potential preventability of cancer that can be achieved by primary prevention may be even higher than our estimates, especially considering other factors in the wider population, including occupational exposures, infectious agents, certain behaviors (e.g., post-menopausal estrogen use), additional dietary factors, and early-life exposures.^37,38

It may be argued that due to health professional background, participants in our cohorts, especially those in the low-risk group, may be more health conscious and accessible to cancer screening and better treatment; and therefore our PARs may have been overestimated. However, we did not find substantial difference between the two risk groups of our cohorts in the uptake of screening. For breast cancer, after standardized to the 2000 US population, the rate of mammographic screening within the past two years was 78% in the high-risk group and 83% in the low-risk group in 2010. For colorectal cancer, the age-standardized rates in 2010 of ever having lower endoscopic screening in the two groups were 63% and 70% in women, and 68% and 75% in men, respectively. Although direct comparison to the US data is difficult because of the differences in assessment methods and timeframes, these estimates are generally consistent with the overall screening uptake in the US,^39,40 suggesting that screening is unlikely to have a substantial influence on our PAR estimates. For cancer therapy, the similar PAR estimates for cancer mortality (67%) and incidence (61%) in men argues against a strong influence of better therapy in our low-risk group relative to the general population, as the reduction in cancer mortality was largely due to a reduction in incidence. In women, the PAR for mortality (59%) was substantially higher than for incidence (41%). Two main factors may have contributed to this difference. First, smoking has a predominant effect that results in a much higher PAR for incident lung cancer than for breast cancer (85% vs. 15%). Yet, despite its low incidence in women, lung cancer is much more fatal than breast cancer, thus contributing relatively greater to the PAR for mortality than for incidence. Indeed, after excluding lung and breast cancers from total carcinoma, we obtained similar PARs for incidence and mortality in women (50% vs. 48%). Secondly, the PAR for breast cancer was much higher for mortality (45%) than for incidence (15%), partly because mammographic screening actually increases cancer incidence due to detection of early or indolent cancer.⁴¹ Additionally, some risk factors, such as obesity and physical inactivity, may influence survival by causing more aggressive cancers, increasing cancer progression, or making cancer more difficult to diagnose early and treat.^42–44 Furthermore, the lifestyle risk factors lower incidence of comorbidities, such as cardiovascular disease and diabetes, which may affect cancer prognosis directly or indirectly (e.g., by limiting aggressive therapy).⁴⁵

Finally, we only included whites in our PAR estimates, which may not be generalizable to other ethnic groups. However, all of the considered factors have been established as risk factors in diverse ethnic groups, although there could be differences in the magnitudes of the associations.

In conclusion, we found that a substantial proportion of cancer cases and even more deaths in the US whites may be prevented if all individuals quit smoking, avoided heavy alcohol drinking, maintained a BMI of 18.5–27.5 kg/m², and exercised in a moderate intensity for at least 150 minutes or in a vigorous intensity for at least 75 minutes every week. These findings reinforce the predominant importance of lifestyle factors in determining cancer risk. Therefore, primary prevention should remain a priority for cancer control.

Abbreviations

AHEI

Alternate Healthy Eating Index

BMI

body mass index

CDC

Centers for Disease Control and Prevention

CI

confidence interval

FFQ

food frequency questionnaire

HPFS

Health Professionals Follow-up Study

MET

metabolic-equivalent

NHANES

National Health and Nutrition Examination Survey

NHS

Nurses’ Health Study

PAR

population attributable risk

SEER

Surveillance, Epidemiology, and End Results