Colorectal Cancer Epidemiology in the Nurses’ Health Study

Abstract

Objectives. To review the contribution of the Nurses’ Health Study (NHS) to identifying risk and protective factors for colorectal adenomas and colorectal cancer (CRC).

Methods. We performed a narrative review of the publications using the NHS between 1976 and 2016.

Results. Existing epidemiological studies using the NHS have reported that red and processed meat, alcohol, smoking, and obesity were associated with an increased risk of CRC, whereas folate, calcium, vitamin D, aspirin, and physical activity were associated with decreased risk of CRC. Moreover, modifiable factors, such as physical activity, vitamin D, folate, insulin and insulin-like growth factor binding protein-1, and diet quality, were identified to be associated with survival among CRC patients. In recent years, molecular pathological epidemiological studies have been actively conducted and have shown refined results by molecular subtypes of CRC.

Conclusions. The NHS has provided new insights into colorectal adenomas, CRC etiology, and pathogenic mechanisms. With its unique strengths, the NHS should continue to contribute to the field of CRC epidemiology and play a major role in public health.

Colorectal cancer (CRC) is the second most commonly diagnosed cancer in women and the third in men worldwide.¹ In 2012, an estimated 1.36 million new cases (women: 614 000; men: 746 000) of CRC were diagnosed, which accounted for 9.7% of total cancers, excluding nonmelanoma skin cancer.¹ The rates vary more than 10 times across the world; high-income countries have approximately 2.5 times higher rates than do low-income countries. Moreover, CRC is the third leading cause of cancer death in women and the fourth in men globally, with the combined number of total deaths reaching 694 000 (8.5% of total cancer deaths).¹ The number of cancer survivors has also grown rapidly over the past several decades. Research on CRC has drawn much attention.

Many epidemiological studies, including the Nurses’ Health Study (NHS), have been conducted to provide evidence for CRC prevention and for improving survival among patients with CRC. We have briefly summarized the key findings from the NHS, a pioneering large prospective cohort study (Tables 1 and 2).

TABLE 1—

Summary of Selected Studies: Nurses’ Health Study, 1976–2016

Study	Study Cohort (Study Years)	Total No. (No. Cases)	Exposure	Outcome	RR (95% CI)	Adjustments
Fat/Fiber
Willett et al.2	NHS (1980–1986)	88 751 (150)	Animal fat	CC	1.89 (1.13, 3.15)	Age, total energy intake
			Quintile 5 vs 1
			Red meat		2.49 (1.24, 5.03)
			≥ 1/d vs < 1/mo
			Processed meat		1.21 (0.53, 2.72)
			≥ 1/d vs < 1/mo
			Cereal fiber		0.74 (0.43, 1.21)
			Quintile 5 vs 1
			Fruit fiber		0.62 (0.37, 1.05)
			Quintile 5 vs 1
			Vegetable fiber		1.07 (0.65, 1.76)
			Quintile 5 vs 1
Michels et al.3	NHS (1984–2000)	76 947 (919)	Cereal fiber	CRC	0.94 (0.79, 1.11)	Age, period, family history of CRC, history of sigmoidoscopy or colonoscopy, height, BMI, PA, aspirin, smoking, multivitamin, total energy intake, alcohol, dietary folate, red meat, processed meat, glycemic load, calcium, methionine, menopausal status, hormone use
			Quintile 5 vs 1
			Fruit fiber		0.87 (0.73, 1.04)
			Quintile 5 vs 1
			Vegetable fiber		1.05 (0.90, 1.23)
			Quintile 5 vs 1
Bernstein et al.4	NHS (1980–2010)	87 108 (1735)	Processed meat Per 1 serving/d	CRC	Pooled HR (95% CI) 1.15 (1.01, 1.32)	Age, 2-follow-up cycle, family history of CRC, history of endoscopy, smoking, BMI, PA, multivitamin, menopausal status and hormone use, aspirin, total energy intake, alcohol, folate, calcium, vitamin D, fiber
	HPFS (1986–2010)	47 389 (996)
				Proximal CC	0.99 (0.79, 1.24)
				Distal CC	1.36 (1.09, 1.69)
Folate
Giovannucci et al.5	NHS (1980–1990)	15 984 (564)	Methyl availability	CRA	0.66 (0.46, 0.95)	Age, family history of CRC, indications for endoscopy, history of endoscopy, total energy intake, saturated fat, fiber, BMI
			Total folate
			Quintile 5 vs 1
			Alcohol intake 2 drinks/d vs 0		1.84 (1.19, 2.86)
			High alcohol and low folate vs low alcohol and high folate		2.71 (1.61, 4.58)
Giovannucci et al.6	NHS (1980–1994)	88 756 (442)	Total folate > 400 vs < 200 μg/d	CC	0.69 (0.52, 0.93)	Age, family history of CRC, aspirin, smoking, BMI, PA, red meat, alcohol, methionine, fiber
Chen et al.7	NHS (1989–1994)	970 (257)	MTHFR	CRA	Total CRA	Age family history, smoking, BMI, folate, methionine, alcohol, fiber, saturated fat
			Genotype: val/val vs val/ala or ala/ala		1.35 (0.84, 2.17)
					Small CRA
					1.36 (0.76, 2.45)
					Large CRA
					1.32 (0.66, 2.66)
Wu et al.8, a	NHS, HPFS (1996–2004)	672 (9134)	Folic acid supplementation 1 mg/d vs placebo	Recurrent CRA	0.87 (0.65, 1.16)	Age, gender, length of trial, time between start of trial and last endoscopy
Lee et al.9	NHS, HPFS (1980–2004)	135 151 (2299)	Total folate > 800 vs < 250 μg/d	CRC	0.69 (0.51, 0.94)	Age, calendar year, smoking, PA, aspirin, height, BMI, family history of CRC, menopausal status and hormone therapy, history of endoscopy, red meat, alcohol, calcium, total energy intake
		79 652 (5655)	Total folate > 800 vs < 250 μg/d	CRA	0.68 (0.60, 0.78)	Further adjusting for recent endoscopy year, indications for endoscopy
Cho et al.10, b	NHS (1989–2010)	1825 (618)	Plasma level of unmetabolized folic acid	CRC	1.03 (0.73, 1.46) 1.12 (0.81, 1.55)	Age, date of blood draw, gender, race, height, fasting status, smoking, BMI, PA, family history of CRC, history of screening, alcohol, red and processed meat, vitamin D, calcium, aspirin
	HPFS (1993–2010)		< 0.5 nmol/L vs 0 ≥ 0.5 nmol/L vs 0
Calcium and vitamin D
Kampman et al.11	NHS (1980–1988)	8935 (350)	Total vitamin D	CRA	0.68 (0.41, 1.13)	Age, total energy intake, BMI, alcohol, folate, saturated fat, fiber, indications for endoscopy, history of endoscopy, family history of colon cancer
			Quintile 5 vs 1
Martinez et al.12	NHS (1980–1992)	89 448 (501)	Total vitamin D	CRC	0.42 (0.19, 0.91)	Age, BMI, PA, family history of CRC, aspirin, smoking, red meat, alcohol
			Quintile 5 vs 1
Wu et al.13	NHS (1980–1996)	87 998 (626)	Total calcium > 1250 vs ≤ 500 mg/d	Distal CC	Pooled RR (95% CI) 0.65 (0.43, 0.98)	Age, family history, BMI, PA, smoking, aspirin, red meat, alcohol, postmenopausal hormone use, menopausal status
	HPFS (1986–1996)	47 344 (399)
Feskanich et al.14, b	NHS (1989–2000)	576 (193)	Plasma 25(OH)D Quintile 5 vs 1	CRC	0.53 (0.27, 1.04)	Age, time of blood draw, BMI, PA, smoking, menopausal status, HRT use, aspirin, family history of CRC, calcium, folate, methionine, retinol, red meat, alcohol
Wu et al.15, b	NHS (1989–2000)	576 (193)	Plasma 25(OH)D Quintile 5 vs 1	CRC	Pooled OR (95% CI) 0.66 (0.42, 1.05)	Age, time of blood draw, BMI, PA, smoking, aspirin, family history of CRC, calcium, folate, retinol, red and processed meat, alcohol (NHS additionally included menopausal status and postmenopausal hormone use)
	HPFS (1993–2002)	535 (179)		CC	0.54 (0.34, 0.86)
Smoking
Giovannucci et al.16	NHS (1976–1990)	12 143 (564)	Smoking pack-years accumulated within the past 20 y	CRA	Small adenoma	Age, saturated fat, fiber, folate, alcohol, BMI, family history of CRC, pack-years of cigarettes smoked in the past
					1.45 (1.25, 1.68)
					Large adenoma
					1.31 (1.17, 1.47)
		118 334 (586)	Among women who started smoking > 10 cigarettes/d	CRC	After 35–39 y follow-up
					1.47 (1.07, 2.01)
					After 40–44 y follow-up
					1.63 (1.14, 2.33)
					After 45 y
					2.00 (1.14, 3.49)
Kenfield et al.17	NHS (1980–2004)	104 519 (578)	Smoking	CRC death	Current smoker	Age, follow-up period, history of hypertension, diabetes, high cholesterol levels, BMI, change in weight, alcohol, PA, oral contraceptives use, postmenopausal estrogen therapy use and menopausal status, parental history of disease, age at starting smoking, red and processed meat, calcium, folate, aspirin
			Current vs never smoker		1.63 (1.29, 2.05)
			Former vs never smoker		Former smoker
					1.23 (1.02, 1.49)
Energy balance
Giovannucci et al.18	NHS (1986–1992)	13 057 (439)	Total PA (MET-h/wk) Quintile 5 vs 1	CRA	0.58 (0.40, 0.86)	Age, family history of CRC, history of endoscopy, smoking, aspirin, animal fat, fiber, alcohol, folate, methionine
			WC, WHR 1 quintile increment		WC
					1.55 (1.09, 2.21)
					WHR
					1.55 (1.08, 2.21)
Wei et al.19, b	NHS (1989–2000)	532 (182)	C-peptide	CC	1.76 (0.85, 3.63)	Age, date of blood draw, fasting status, BMI, PA, smoking, alcohol, family history of CRC, aspirin, history of screening, menopausal status, postmenopausal hormones
			Quartile 4 vs 1
			IGFBP-1		0.28 (0.11, 0.75)
			Quartile 4 vs 1
			IGF-1/IGFBP-3		2.82 (1.35, 5.88)
			Quartile 4 vs 1
Wei et al.20, b	NHS (1989–1998)	760 (380)	C-peptide Quartile 4 vs 1	CRA	1.63 (1.01, 2.66)	Age, period of and indications for endoscopy, date of blood draw, BMI, PA, smoking, alcohol, family history of CRC, aspirin, menopausal status, HRT use
Survival
Meyerhartdt et al.21	NHS (1986–2004)	573 (72)	Postdiagnosis PA ≥ 18 vs < 3 MET h/wk	CRC death	0.39 (0.18, 0.82)	Age, BMI, stage of disease, grade of tumor differentiation, location of primary tumor, year of diagnosis, chemotherapy, time from diagnosis to PA measurement, change in BMI before and after diagnosis, smoking
		573 (121)		Overall death	0.43 (0.25, 0.74)
Ng et al.22, b	NHS (1989–2005)	304 (96)	Plasma 25(OH)D Quartile 4 vs 1	CRC death Overall death	Pooled HR (95% CI) 0.61 (0.31, 1.19) 0.52 (0.29, 0.94)	Age, season of blood draw, gender, stage of disease, grade of tumor differentiation, location of primary tumor, year of diagnosis, BMI at diagnosis, postdiagnosis PA
	HPFS (1993–2005)	304 (123)
Wolpin et al.23, b	NHS (1989–2005)	301 (95)	Plasma folate Quintile 5 vs 1	CRC death Overall death	Pooled HR 0.42 (0.20, 0.88) 0.46 (0.24, 0.88)	Age, stage of disease, histologic differentiation, chemotherapy, tumor location, period of diagnosis, BMI, PA, smoking, aspirin, alcohol, total vitamin D, postmenopausal hormone use
	HPFS (1993–2005)	301 (122)
Wolpin et al.24, b	NHS (1989–2004)	373 (108)	C-peptide Quartile 4 vs 1 IGFBP-1 Quartile 4 vs 1	Overall death	2.11 (1.06, 4.21)	Age, gender, stage of disease, histologic differentiation, tumor location, period of diagnosis, time between last meal and plasma collection, chemotherapy, smoking, aspirin, alcohol, total vitamin D, postmenopausal hormone
	HPFS (1993–2004)				0.44 (0.24, 0.81)
Fung et al.25	NHS (1986–2010)	1201 (162)	AHEI-2010	CRC death	0.72 (0.43, 1.21)	Age, PA, BMI, weight change, stage of disease, chemotherapy, smoking, total energy intake, colon or rectal cancer, stage of disease, date of CRC diagnosis
			Quintile 5 vs 1
		1201 (435)		Overall death	0.71 (0.52, 0.98)
Molecular pathogenic epidemiology
Liao et al.26	NHS (1980–2011)	964 (190)	Aspirin Postdiagnosis regular use vs no use	CRC death Overall death	Pooled HR (95% CI) PIK3CA mutation 0.18 (0.06, 0.61) 0.54 (0.31, 0.94)	Age, gender, stage of disease, BMI, year of diagnosis, time from diagnosis to first measurement of aspirin use after diagnosis, regular use or nonuse of aspirin before diagnosis, tumor location, tumor differentiation, microsatellite instability status, CIMP, KRAS mutation, BRAF mutation, LINE-1 methylation, and the presence or absence of PTGS2 expression
	HPFS (1986–2011)	964 (395)			Wild-type PIK3CA 0.96 (0.69, 1.32) 0.94 (0.75, 1.17)
Nishihara et al.27	NHS (1980–2008)	134 204 (714)	Smoking duration of cessation (10–19, 20–39, ≥ 40 y) vs current smoker	CRC	CIMP-high CRC 0.53 (0.29, 0.95) 0.52 (0.32, 0.85) 0.50 (0.27, 0.94)	Age, gender, BMI, family history of CRC, aspirin, PA, alcohol, total energy intake, red meat
	HPFS (1986–2008)
					CIMP-low CRC
					1.07 (0.81, 1.42)
					0.98 (0.77, 1.26)
					0.95 (0.69, 1.32)

Note. 25(OH)D = 25-hydroxyvitamin D; AHEI-2010 = alternative healthy eating index-2010; BMI = body mass index; CC = colon cancer; CI = confidence interval; CRA = colorectal adenoma; CIMP = CpG island methylator phenotype; CRC = colorectal cancer; HPFS = health professional follow-up study; HR = hazard ratio; HRT = hormone replacement therapy; IGF = insulin-like growth factor; IGFBP = insulin-like growth factor binding protein; LINE-1 = long interspersed nuclear element 1; MET = metabolic equivalent task; MTHFR = methylenetetrahydrofolate reductase; NHS = Nurses’ Health Study; OR = odds ratio; PA = physical activity; RR = rate ratio; WC = waist circumference; WHR = waist to hip ratio.

^aRandomized controlled trial.

^bNested case–control study.

TABLE 2—

Initial Findings on Factors Related to Colorectal Cancer Risk and Survival: Nurses’ Health Study, 1976–2016

Publication Year	Factors	Association
CRC risk
1990	Red meat	Positive
1990	Fiber	Null
1993	Folate	Inverse
1993	Alcohol	Positive
1994	Calcium and vitamin D	Inverse
1994	Smoking	Positive
1995	Aspirin	Inverse
1996	Physical activity	Inverse
1996	BMI	Positive
	Waist circumference	Positive
2013	Smoking cessation	Inverse (CIMP-high CRC)
		Null (CIMP-low CRC)
Mortality in CRC patients
2006	Physical activity	Inverse
2008	Vitamin D	Inverse
2008	Folate	Inverse
2009	Insulin	Positive
	IGFBP-1	Inverse
2012	Aspirin	Inverse (PIK3CA mutation)
		Null (wild type)
2014	Diet quality	Inverse

Note. BMI = body mass index; CIMP = CpG island methylator phenotype; CRC = colorectal cancer; IGFBP-1 = insulin-like growth factor binding protein-1.

DIETARY FAT

In 1990, little was known about the causes of CRC. One of the prevailing dietary hypotheses was that dietary fat, especially from animal sources, might increase CRC risk. This hypothesis arose from the results of studies examining the correlation between per capita consumption of animal fat and national disease rates across countries and of case–control studies comparing recalled past diet between individuals with and without CRC. These studies generally did not account for total energy intake, complicating the interpretation of their positive findings.

Moreover, the case–control studies were potentially prone to recall and selection bias. A few cohort studies were available, but they were limited by the low number of cases and the noncomprehensive dietary assessments. In 1990, Willett et al.,² whose study was among the first prospective cohort studies on diet and cancer, showed that higher consumption of energy-adjusted animal fat was positively associated with colon cancer risk (relative risk [RR] for highest vs lowest quintile: 1.89; 95% confidence interval [CI] = 1.13, 3.15). The association was limited largely to fat from red and processed meat; women with daily consumption of red meat (i.e., beef, pork, or lamb) had approximately a 2.5 times higher risk compared with women consuming these items less than once a month.

Moreover, significant positive trends were shown with higher processed meat consumption (trend P = .04). A later study using both the NHS and the Health Professional Follow-Up Study (HPFS) cohorts showed that processed red meat was positively associated with the risk of CRC, particularly with distal colon cancer.⁴ The association did not change when dietary cholesterol and saturated fat were adjusted for. Other main sources of fat (e.g., dairy, chicken, and vegetable oil) were not associated with the risk of CRC. In October 2015, the International Agency for Research on Cancer declared that the consumption of red meat was probably carcinogenic to humans and that processed meat was carcinogenic to humans according to the scientific evidence.²⁸

FIBER

Fiber was hypothesized to protect against CRC in the late 1960s. The NHS first examined the association between dietary fiber from diverse sources and colon cancer in 1990.² Although a suggestive inverse association was found specifically for fruit fiber (RR for highest vs lowest quintile = 0.62; 95% CI = 0.37, 1.05), a subsequent study reported an either null or weak inverse association between dietary fiber and CRC or colorectal adenoma (CRA) risk.^3,29

Recently, a meta-analysis of observational studies found a significant inverse association. However, an inverse association between fiber and CRC risk was largely attenuated or even disappeared when adjusted for additional confounders in several observational studies, including the NHS. Also, no significant results for fiber were found in randomized controlled trials studying recurrent CRA. More studies are warranted to examine the association between fiber and colorectal neoplasia that account for fiber type (food vs supplementary), dietary fiber sources (fruits, vegetables, and grains), and intestinal microflora profiles, which may be influenced by fiber or may modify the effect of fiber.

FOLATE

Folate plays an essential role in maintaining the integrity of DNA synthesis and methylation. A potential role of folate in colorectal carcinogenesis was hypothesized on the basis of studies showing a reduced risk of colorectal neoplasia with an increased consumption of fruits and vegetables, which are primary sources of folate in some settings. In 1993, an initial prospective cohort study on the topic discovered that a high intake of folate was inversely associated with CRA risk (RR for highest vs lowest quintile = 0.66; 95% CI = 0.46, 0.95) in the NHS.⁵ The major sources of folate were multivitamins, fortified breakfast cereals, and fortified flour products. The association of CRA risk with folate was independent of fruit and vegetable consumption. In addition, alcohol is known to antagonize folate absorption and function. In the study, high consumption of alcohol (2 drinks/day) was associated with an increased risk of CRA compared with no consumption of any alcohol (RR = 1.84; 95% CI = 1.19, 2.86).⁵

Moreover, those with a high-alcohol and low-folate intake had a 2.7 times higher risk of CRA than did those with a low-alcohol and high-folate intake. In 1 of the earliest studies examining genetic polymorphisms in relation to cancer risk, results from the NHS indicated that a variant of the methylenetetrahydrofolate reductase gene, which affects the metabolism of folate, predicted the risk of CRA—offering further support for a role of folate in colorectal carcinogenesis.⁷

In 1998, a subsequent NHS examined the role of folate on CRC risk.⁶ Consistent with findings on CRA, a significantly reduced risk was observed with higher total folate intake from both dietary and supplemental sources (RR for > 400 vs < 200 μg/day = 0.69; 95% CI = 0.52, 0.93). Furthermore, when multivitamins containing folic acid were examined alone, significant results were not shown before 15 or more years of use, but a substantial reduction in the risk of CRC was present after 15 years of use (RR = 0.25; 95% CI = 0.13, 0.51). This finding suggests that a long period of folate intake is required to exert a benefit on CRC risk. A pooled analysis of 13 prospective studies published in 2010 also provided supporting evidence that higher folate intake is associated with a lower risk of colon cancer. However, the results from randomized controlled trials were weaker. In a randomized controlled trial that recruited participants from the NHS and the HPFS, folic acid supplementation did not reduce CRA recurrence, although a potential benefit was suggested for those with low-folate status at baseline.⁸

One of the strengths of the NHS is repeated measures of diet, which allow the assessment of time lags between exposure and the risk of CRC. A study combining NHS and HPFS published in 2011 found that a long induction period was required to observe the benefit of adequate folate intake on CRC risk.⁹ Total folate intake 12 to 16 years before diagnosis was associated with a reduced risk of CRC (RR for ≥ 800 vs < 250 μg/day = 0.69; 95% CI = 0.51, 0.94), whereas total folate intake close to diagnosis (4–8 years) was most strongly associated with a reduced risk of CRA (odds ratio [OR] for ≥ 800 vs < 250 μg/day = 0.68; 95% CI = 0.60, 0.78). The long latency period suggests that folate may act on the initiation or early development of colorectal carcinogenesis.

Recently, there have been growing concerns that excessive intake of folate may actually increase cancer risk. For example, a temporary increase in CRC rates in the United States was observed immediately after the implementation of folic acid fortification in 1997. Cho et al., using NHS data, found that prediagnostic plasma levels of unmetabolized folic acid, which increases when folic acid intake is too high to be fully metabolized, was not associated with CRC risk.¹⁰ This result provides strong reassurance that folic fortification has not caused an increase in CRC.

CALCIUM AND VITAMIN D

Calcium and vitamin D intakes were hypothesized to be protective factors against CRC on the basis of experimental studies and ecological studies; however, case–control studies had inconsistent results. In 1996, the NHS found that high long-term average intake of total vitamin D was associated with a 58% reduced risk of CRC.¹² Vitamin D can be obtained from dietary and supplemental sources and produced by the skin when exposed to ultraviolet B radiation. Plasma 25-hydroxyvitamin D (25(OH)D) is considered a comprehensive biomarker of vitamin D status. The NHS collected and stored blood samples from a study population subset in 1989. After up to 11 years of follow-up since the blood collection, in 2004, a nested case–control study in the NHS showed, for the first time, a statistically significant inverse association between the prediagnostic 25(OH)D level and CRC risk (trend P = .02).¹⁴ The pooled results of NHS and HPFS reported in 2007 further supported an inverse association between plasma 25(OH)D level and CRC incidence.¹⁵

The initial NHSs, which had a short-term follow-up period, did not support the inverse association between calcium and the risk of CRA and CRC.^11,12 However, an examination of the combined NHS and HPFS cohorts, which had a long-term follow-up period (NHS: 1980–1996; HPFS: 1986–1996), found a significant inverse association between total calcium intake and distal colon cancer risk (RR for > 1250 vs ≤ 500 mg/day = 0.65; 95% CI = 0.43, 0.98).¹³ This result demonstrates the importance of continuing follow-up in cohort studies to accumulate more power to detect moderate associations. An inverse association between calcium and CRC and adenomas has now been observed consistently in prospective studies. The World Cancer Research Fund International/American Institute for Cancer Research has declared diets high in calcium to be a probable protective factor for CRC.³⁰ Randomized trials to date have been inconsistent regarding calcium and inadequate to examine the role of vitamin D owing to low doses and short follow-up periods. Some large ongoing randomized trials testing higher doses with longer follow-up periods may provide invaluable evidence to assess the vitamin D and CRC relationship.

SMOKING

Currently, smoking is among the strongest established risk factors for overall cancers. However, in epidemiological studies before the 1990s, a positive association between smoking and CRC was shown only among men; the results were mostly null or inconsistent among women. The NHS was the first study to show that a long induction period of several decades is required for smoking to increase the risk of CRC, which could explain the null findings from the previous studies with limited follow-up.¹⁶

Moreover, among women who had started smoking more than 10 cigarettes per day, cigarette smoking was not associated with an increased risk of CRC until 35 years after smoking was initiated. Accounting for a 35-year time lag, a monotonic dose–response relationship was shown between pack-years of cigarettes smoked and CRC. Similar results were found for CRC mortality.¹⁷ The hypothesis of a long time lag between smoking and CRC risk was supported by subsequent epidemiological studies and is now acknowledged in the surgeon general’s report.³¹ In 2011, the International Agency for Research on Cancer classified tobacco smoking as a carcinogenic agent for colon and rectum sites, with sufficient evidence in humans.³²

ENERGY BALANCE

In the late 1900s, epidemiological studies began to investigate the potential role of physical activity and body mass index (BMI; defined as weight in kilograms divided by the square of height in meters) in colorectal neoplasia. Interestingly, physical activity and excess adiposity were consistently associated with colon cancer risk in men, but not women, suggesting a potential gender difference. However, in 1996, the NHS was the first prospective study to show a significant inverse association between physical activity and CRA among women, with the most physically active women at a 40% lower risk.¹⁸ The association was stronger for large adenomas, the proximal precursors to CRC.

With regard to obesity, an increased risk of large CRA was shown when comparing women with a BMI of 29 or higher versus those with a BMI of less than 21 (RR = 2.21; 95% CI = 1.18, 4.16). Interestingly, the NHS first found that abdominal adiposity measured by waist circumference and waist to hip ratio was associated with a risk of large CRA.¹⁸ Furthermore, women with a high waist to hip ratio and a high BMI had a greater risk of large adenoma (RR = 1.99; 95% CI = 0.98, 4.05) than did those with a low waist to hip ratio and a high BMI (RR = 1.35; 95% CI = 0.61, 2.97). For the colon cancer outcome, similar results were found in 1997.³³

The demonstration of physical inactivity and obesity (both overall and abdominal) as risk factors for CRA and CRC was the basis of the novel hypothesis that hyperinsulinemia may increase CRC risk. Abdominal adiposity and physical inactivity are primary determinants of insulin resistance and hyperinsulinemia. Later studies in the NHS as well as in other populations have supported this hypothesis. In 2005 and 2006, nested case–control studies were conducted within the NHS subcohort who had provided blood samples in 1989 to test the hyperinsulinemia hypothesis.^19,20 After more than 10 years of follow-up, a direct association was observed between circulating levels of C-peptide, an indicator for insulin secretion, and incident colon cancer (RR for highest vs lowest quartile = 1.76; 95% CI = 0.85, 3.63).

Furthermore, high levels of fasting insulin-like growth factor binding protein (IGFBP)-1, which reflect low insulin, were strongly inversely associated with colon cancer (RR for highest vs lowest quartile = 0.28; 95% CI = 0.11, 0.75), and the insulin-like growth factor (IGF)-1 to IGFBP-3 molar ratio was associated with colon cancer (RR for highest vs lowest quartile = 2.82; 95% CI = 1.35, 5.88). In addition, higher C-peptide was significantly associated with a 63% increased risk of CRA, even after further adjusting for BMI and physical activity. These findings, also supported in other populations, strongly indicated that alterations in insulin and IGFs underlie the association between physical inactivity and central adiposity and colon cancer risk.

Currently, total body and abdominal fatness are considered convincing risk factors, whereas physical activity is recognized as a protective factor against CRC by many national and international cancer organizations such as the World Cancer Research Fund International/American Institute for Cancer Research,³⁰ the International Agency for Research on Cancer,³⁴ and the National Cancer Institute.³⁵ A consensus conference cosponsored by both the American Cancer Society and the American Diabetes Association highlighted hyperinsulinemia as a potential mechanism underlying the association between obesity and physical activity with colorectal as well as other cancers.³⁶

SURVIVAL

In addition to many important findings related to CRC prevention, the NHS was among the first studies to examine potential modifiable factors for survival among patients with CRC. In the early 2000s, little was known about risk factors and protective factors that could improve the survival of CRC patients. Because of continuing follow-up and updating of data in the NHS, the relation between lifestyle factors and CRC-specific and overall mortality can be examined. Because postdiagnostic exposures are likely to be associated with and thus potentially confounded by prediagnosis exposure, an advantage of the NHS is the ability it provides to examine the independent associations of pre- and postdiagnostic exposures in relation to survival.

In 2006, Meyerhardt et al. examined the association between postdiagnosis physical activity and CRC-specific and overall mortality among women with stages I–III CRC.²¹ They found that women engaging in 18 or more metabolic equivalent task-hours per week of physical activity had a significantly lower risk for CRC-specific mortality (hazard ratio [HR] = 0.39; 95% CI = 0.18, 0.82) and overall mortality (HR = 0.43; 95% CI = 0.25, 0.74) than did those engaging in fewer than 3 metabolic equivalent task-hours per week of physical activity. This finding suggests that postdiagnosis physical activity may confer additional benefit beyond the standard cancer therapy for patients with CRC. In 2009, Wolpin et al. found that prediagnosis plasma levels of C-peptide and IGFBP-1 were associated with mortality in patients with CRC, with adjusted HRs of overall mortality comparing the extreme quartiles of C-peptide and IGFBP-1 of 2.11 (95% CI = 1.06, 4.21) and 0.44 (95% CI = 0.24, 0.81), respectively.²⁴

Some of the dietary factors known or suspected to influence CRC risk, such as vitamin D, folate, and overall dietary quality score, were examined in relation to survival. In 2008, Ng et al. first reported a significant inverse association between prediagnosis levels of circulating 25(OH)D and mortality among CRC patients, with the adjusted HR comparing the extreme quartiles being 0.52 (95% CI = 0.29, 0.94).²² Wolpin et al. provided evidence to ameliorate the concern that higher folate levels around the time of CRC diagnosis may enhance carcinogenesis.²³ Participants in the highest quintile had a greater than 50% reduced risk of CRC-specific mortality and overall mortality than did participants in the lowest quintile of plasma folate levels.

Considering the increased public consumption of folate with mandatory folic acid fortification, this finding disputes the hypothesized harmful effect of higher folate intake on colorectal neoplasia progression. The NHS was also the first study to examine mortality among CRC patients in relation to the Alternative Healthy Eating Index-2010, a measure of diet quality used to access patient compliance with the US dietary guidelines²⁵; results were published in 2014. Women with a healthier dietary pattern, as indicated by a higher Alternative Healthy Eating Index-2010 score, have a decreased risk of overall mortality (HR = 0.71; 95% CI = 0.52, 0.98). The inverse association was primarily explained by moderate alcohol intake (relative to no or high intakes) and lower consumption of sugar-sweetened beverages and fruit juices, among all components of the Alternative Healthy Eating Index-2010.

GENOME-WIDE ASSOCIATION STUDY

Subsequent to our increased understanding of the importance of genetic variants in CRC risk, the Genetics and Epidemiology of Colorectal Cancer Consortium, which includes the NHS, has reorganized to accelerate finding genetic variants related to CRC and their interaction with environmental risk factors. The pooling of multiple studies increases statistical power. For instance, Hsu et al. showed that models to determine the risk of CRC could be significantly improved by incorporating information on CRC risk alleles.³⁷

Moreover, there have been studies examining gene–environment interactions related to CRC risk. Figueiredo et al. found that the association between processed meat consumption and CRC was modified by genotypes of rs4143094.³⁸ Similarly, Nan et al. detected significant interactions between aspirin or nonsteroidal anti-inflammatory drug use and 2 single nucleotide polymorphisms (rs2965667 and rs10505806).³⁹

MOLECULAR PATHOLOGICAL EPIDEMIOLOGY

CRC is a heterogeneous disease caused by a complex interplay between genetic and epigenetic changes. CRC has been largely classified into 3 molecular subtypes: chromosomal instability, microsatellite instability, and the CpG island methylator phenotype (CIMP). Molecular pathological epidemiology (MPE) posits that risk factors may differ for cancers on the basis of their molecular subtype. MPE studies are refining our understanding of various exposures in relation to CRC risk. MPE studies in the NHS have enhanced our understanding regarding the role of aspirin and smoking in colorectal carcinogenesis.

By 1995, a potential benefit of aspirin against CRC was suggested in the NHS and other cohorts. The NHS showed substantially reduced risk of CRC after 10 or more years of regular aspirin use at doses similar to those used for the prevention of cardiovascular disease.⁴⁰ The identification of a 10-year requirement to observe an association was confirmed in randomized controlled trials published 15 years later. Refining the finding, recent MPE studies revealed that aspirin use is effective in preventing only some molecular subtypes of CRC (e.g., BRAF mutation status, WNT/CTNNB1 signaling, and 15-hydroxyprostaglandin dehydrogenase).^41–43 Another study examined the association between postdiagnosis aspirin use and CRC survival according to tumor PIK3CA mutation status.²⁶ Among patients with PIK3CA mutation, regular use of aspirin was strongly associated with both improved cancer-specific survival (RR = 0.18; 95% CI = 0.06, 0.61) and overall survival (RR = 0.54; 95% CI = 0.31, 0.94). However, among patients with wild-type PIK3CA, regular use of aspirin was not associated with cancer-specific survival (HR = 0.96; 95% CI = 0.69, 1.32) or overall survival (HR = 0.94; 95% CI = 0.75, 1.17). This finding suggests the importance of targeted interventions when considering the chemopreventive potential of aspirin in CRC survival in the future.

In 2013, the first MPE study to examine the relationship between duration of smoking cessation and CRC risk according to molecular subtypes was conducted.²⁷ Among those who quit smoking, time since smoking cessation was significantly associated with a decreased risk of CIMP-high CRC (trend P = .001) but not of CIMP-low CRC (trend P = .25) compared with those currently smoking. This study helped confirm that CRC risk could be reduced on smoking cessation later in life, which had been a controversial topic. Because CIMP-high CRC represents only about 15% to 20% of the total, this finding was obscured in studies that evaluated only total CRC and did not examine subtypes separately.

CONCLUSIONS

Since its initiation in 1976, the NHS has provided important evidence on reducing the burden of CRC. With one of the earliest and largest prospective cohorts, the NHS has greatly contributed to initial findings of risk and protective factors associated with CRC incidence and survival among patients with CRC. The NHS has identified or confirmed red and processed meat, alcohol, smoking, and obesity as risk factors, and folate, calcium, vitamin D, and aspirin intake and physical activity as protective factors. In the NHS, it was found that optimizing exposure to some of these modifiable protective factors had the potential to prevent an estimated 43% of colon cancer.⁴⁴ Through the concurrent study of adenomas and the availability of multiple assessment of exposures over time, the NHS has also provided critical information on the timing of the exposure in relation to the risk of CRC. Moreover, the NHS is among the first studies to have identified modifiable factors for survival in patients with CRC.

Subsequent studies from other cohorts and populations have largely confirmed these findings from the NHS. Currently, findings from the NHS have contributed to convincing evidence that CRC burden can be reduced via lifestyle and dietary modifications.⁴⁵ In recent years, there has been an increasing number of MPE studies of the NHS data; these have provided new insights into the causes of CRC. Unquestionably, the findings from NHS have helped not only establish the field of CRC epidemiology but also make policy recommendations to improve public health. The unique strengths and ongoing development of the NHS cohort should allow scientists to continue novel and valuable research to prevent CRC and to improve quality of life in CRC patients.