Associations Between Anthropometry, Cigarette Smoking, Alcohol Consumption, and Non-Hodgkin Lymphoma in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial

Abstract

Prospective studies of lifestyle and non-Hodgkin lymphoma (NHL) are conflicting, and some are inconsistent with case-control studies. The Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial was used to evaluate risk of NHL and its subtypes in association with anthropometric factors, smoking, and alcohol consumption in a prospective cohort study. Lifestyle was assessed via questionnaire among 142,982 male and female participants aged 55–74 years enrolled in the PLCO Trial during 1993–2001. Hazard ratios and 95% confidence intervals were calculated using Cox proportional hazards regression. During 1,201,074 person-years of follow-up through 2006, 1,264 histologically confirmed NHL cases were identified. Higher body mass index (BMI; weight (kg)/height (m)²) at ages 20 and 50 years and at baseline was associated with increased NHL risk (P_trend < 0.01 for all; e.g., for baseline BMI ≥30 vs. 18.5–24.9, hazard ratio = 1.32, 95% confidence interval: 1.13, 1.54). Smoking was not associated with NHL overall but was inversely associated with follicular lymphoma (ever smoking vs. never: hazard ratio = 0.62, 95% confidence interval: 0.45, 0.85). Alcohol consumption was unrelated to NHL (drinks/week: P_trend = 0.187). These data support previous studies suggesting that BMI is positively associated with NHL, show an inverse association between smoking and follicular lymphoma (perhaps due to residual confounding), and do not support a causal association between alcohol and NHL.

Non-Hodgkin lymphomas (NHL), a heterogeneous group of malignant neoplasms arising in lymphoid cells throughout the body, represent the sixth most common cancer in the United States (1). Diffuse large B-cell lymphoma, follicular lymphoma, chronic lymphocytic leukemia (CLL)/small lymphocytic lymphoma (SLL), and plasma cell neoplasms are the most common NHL subtypes (1). NHL incidence and mortality increased steadily during the last half of the 20th century in the United States, with rates approximately doubling in both sexes during this time (1–3). The age-adjusted (2000 US standard population) incidence of NHL in the United States during 2006 was 29.9 per 100,000 person-years (4). The rise in NHL rates was not entirely explained by the acquired immunodeficiency syndrome epidemic, changes in coding practices, or changes in NHL detection and diagnosis (2). The strongest known risk factors for NHL, infectious agents and immune system dysfunction, account for only a small proportion of cases (2).

Previous research suggests that NHL is associated with obesity, cigarette smoking, and alcohol consumption, and effects may vary by NHL subtype. Pooled analyses of results from case-control studies suggest increased risk of diffuse large B-cell lymphoma with obesity (5, 6), increased risk of follicular lymphoma with cigarette smoking (7), and decreased risk of all types of NHL with alcohol consumption (8). Few cohort studies have examined lifestyle factors and NHL risk; results from these studies are conflicting, and some are inconsistent with those from case-control studies. Specifically, some prospective studies have suggested that body mass index (weight (kg)/height (m)²) affects NHL risk in women and men (9–18), whereas others have shown no association (19–23). Cohort studies have not identified an association between smoking status and the risk of NHL overall (24–28), yet some studies have shown an increased risk of follicular lymphoma (24, 25), whereas others observed an inverse association with follicular lymphoma or no association (26, 28). The few cohort studies that have examined alcohol and NHL risk have all shown an inverse association, with conflicting evidence for the effect of alcoholic beverage type (28–31).

Prospective studies of lifestyle factors and NHL have provided conflicting results and have lacked either an adequate sample size or the histologic information required for examining NHL subtype. Therefore, we evaluated the associations between anthropometric factors, smoking, and alcohol consumption and incident NHL and major NHL subtypes in a population-based cohort of older US men and women.

MATERIALS AND METHODS

Study population

The study population for this analysis was derived from the National Cancer Institute's Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, which has been described previously (32). Briefly, 154,910 men and women aged 55–74 years with no prior history of the cancers under study were enrolled during 1993–2001 at 10 centers around the United States in a randomized, controlled trial designed to evaluate the impact of cancer screening on mortality. Participants completed a baseline questionnaire that assessed demographic information and self-reported anthropometric factors and tobacco use. Beginning in 1998, data on alcohol consumption were collected with other dietary data on a food frequency questionnaire (validated via 24-hour recall) from control-arm participants at baseline and from intervention-arm participants during their fourth year of participation. Both questionnaires were self-administered and returned by mail, self-administered in person; or administered by telephone. Starting in 2006, anthropometric factors and smoking behavior were reassessed on a supplemental questionnaire that was returned by mail (32, 33). Follow-up in the PLCO Trial is ongoing.

Of the 154,910 PLCO participants, 149,984 (96.8%) completed the baseline questionnaire. After excluding 6,789 subjects with a history of previous cancer, 99 subjects diagnosed with cancer prior to completing the questionnaire, and 114 subjects with invalid follow-up time (e.g., the start date followed the stop date), we included 142,982 participants (70,905 women and 72,077 men) in our smoking and anthropometry analyses. A total of 116,736 PLCO participants (75.4%) completed the dietary history questionnaire. After excluding 5,221 subjects with invalid questionnaires (e.g., incomplete data or data outside of expected ranges), 4,885 subjects with a history of previous cancer, 4,741 subjects diagnosed with cancer prior to completing the questionnaire, and 76 subjects with invalid follow-up time, we included 101,813 participants (52,280 women and 49,533 men) in our alcohol analyses. Nonmelanoma skin cancer was excluded when considering participants’ cancer history.

For participants who provided responses on both the baseline and supplemental questionnaires, we evaluated changes over time in current body mass index (n = 91,937), height (n = 93,766), weight (n = 93,894), and smoking status (n = 90,543). The mean time between completion of the questionnaires was 9.1 years.

The National Cancer Institute's institutional review board approved the PLCO Trial protocol.

Follow-up and case ascertainment

We followed participants from the date of completion of the baseline or dietary history questionnaire to the earliest of the following: diagnosis with any NHL or any other first primary cancer (excluding in-situ disease and nonmelanoma skin cancer), death, loss to follow-up, or the end of the study period (December 31, 2006). Deaths in the PLCO Trial are ascertained by means of annual questionnaires as well as through linkage with state vital statistics and the National Death Index (32, 33). We considered subjects lost to follow-up at the date of last contact if they were not known to be deceased and were unresponsive to the most recent attempts at contact.

Cancer cases in the PLCO Trial are identified through annually mailed follow-up questionnaires and telephone calls and are confirmed by medical record review at the screening centers. Histologic type is recorded by a certified tumor registrar at each screening center on the basis of collected medical records and pathology reports using the International Classification of Diseases for Oncology (32). We defined a case as any first primary NHL diagnosed prior to censoring, death, or withdrawal from the study. We defined NHL subtypes using International Classification of Diseases for Oncology codes specified by the World Health Organization-based classification of lymphoid neoplasms recommended by the Pathology Working Group of the International Lymphoma Epidemiology Consortium (34, 35).

Exposure data

Anthropometric factors.

Baseline height (in feet and inches) and weight (in pounds) at ages 20 and 50 years and at baseline were solicited on the baseline questionnaire and converted to centimeters and kilograms, respectively, for analysis and for calculation of body mass index at each age. Quartiles of height, weight (at each age), and body mass index (at each age) were created separately for men and women. Body mass index was also analyzed according to World Health Organization categories (<18.5, 18.5–24.9, 25–29.9, and ≥30). We also analyzed the effects of lifetime weight change by examining weight change in 10-year increments from age 20 years to baseline. We conducted analyses in the total population (adjusted for sex) and for each sex separately to evaluate potential sex-specific effects on disease risk.

Smoking.

Smokers were defined as participants with a minimum 6-month regular smoking history, as reported on the baseline questionnaire. Additional data on age at which participants started smoking, current smoking status, age at which participants stopped smoking, and usual number of cigarettes smoked per day (1–10, 11–20, 21–30, 31–40, 41–60, 61–80, or ≥81) were solicited. We computed cumulative lifetime exposure to cigarette smoking (pack-years) on the basis of reported data on packs of cigarettes smoked per day (midpoint of the categories listed above) and years of smoking.

Alcohol consumption.

Data on alcohol consumption were solicited on the PLCO dietary history questionnaire by first asking whether participants had consumed beer, wine/wine coolers, or liquor during the preceding 12-month period. If so, data on frequency and usual serving size (small, medium, or large) were solicited for each beverage type. Data on beer consumption were solicited separately for the summer months versus the remainder of the year. Serving sizes were defined for each beverage type as <1, 1–3, or >3 12-ounce (355-mL) cans for beer; <5 ounces (<148 mL) or <1 glass, 5–12 ounces (148–355 mL) or 2 glasses, or >12 ounces (>355 mL) or >2 glasses for wine or wine coolers; and <1, 1–3, or >3 shots for liquor. Daily alcohol consumption (grams of each beverage) was converted to number of drinks per week using standardized sex- and beverage-specific gram equivalents for the usual serving sizes (see Web Table 1, which is posted on the Journal’s Web site (http://aje.oxfordjournals.org/)). Total ethanol consumption (g/week) was calculated for each beverage type, defining 1 drink as a 12-ounce (355-mL) beer containing 12.8 g of ethanol, a 4-ounce (118-ml) glass of wine containing 11 g of ethanol, or a 1.5-ounce (44-mL) serving of liquor containing 14 g of ethanol.

We analyzed the effect of alcohol consumption on NHL risk according to total number of drinks per week and total grams of ethanol consumed per week. We also evaluated effects of specific alcoholic beverages by analyzing separately the numbers of drinks of beer, wine, and liquor consumed per week. Nondrinkers were defined as participants who consumed less than 1 drink per month. Light drinkers (<1 drink or <11.8 g of ethanol per week) were used as the referent group for all categorical analyses.

Statistical analyses

Cox proportional hazards regression modeling was used to estimate the risks of developing NHL overall and of the major NHL subtypes (diffuse large B-cell lymphoma, follicular lymphoma, CLL/SLL, and plasma cell neoplasms) associated with anthropometric characteristics, smoking, and alcohol consumption. When analyzing a single NHL subtype, other subtypes were censored. Models were adjusted for age (continuous), race/ethnicity (non-Hispanic white vs. other/unknown), sex, and education (high school or less, post-high school training or some college, and college graduation or postgraduate study). Additional adjustment for several dietary factors (total energy intake (kcal/day); g/day of fat, meat, and red meat; and pyramid servings/day of total fruit, citrus fruit/melon/berries, other types of fruit, all vegetables, and green vegetables), as well as mutual adjustment for the 3 lifestyle factors, did not materially (>10%) alter the results; therefore, those factors were excluded from the final models. To evaluate the consistency of the smoking risk estimates in different demographic groups, we also conducted analyses stratified by sex, age, and education.

The proportional hazards assumption was verified graphically for all independent variables. We conducted sensitivity analyses after excluding the first year of follow-up to eliminate possible subclinical prevalent disease at baseline. Tests for linear trend were conducted using continuous variables in the Cox regression models, including all participants for anthropometric characteristics, smokers only for smoking exposure variables, and drinkers only for alcohol consumption variables. All statistical tests were 2-sided.

All analyses were conducted using SAS, version 9.1.3 (SAS Institute, Cary, North Carolina).

RESULTS

During 1,201,074 person-years of follow-up (median, 8.8 years per person) accrued through December 31, 2006, 1,264 histologically confirmed cases of first primary NHL were diagnosed among 70,905 women and 72,077 men (age-adjusted incidence rate using 2000 US standard population: 16.6/100,000 person-years) (Table 1). The most common NHL subtype was CLL/SLL (30.2%), followed by plasma cell neoplasms (19.2%), diffuse large B-cell lymphoma (17.0%), and follicular lymphoma (12.8%). A similar distribution was observed in the subset of cases (n = 700) available for the alcohol analysis. The median age at diagnosis for all NHL cases was 64 years.

Table 1.

Distribution of Cases of Non-Hodgkin Lymphoma^a Identified During Follow-Up of 142,982 Participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, 1993–2006

	Smoking and Anthropometry Analysis		Alcohol Analysis
	No.	%	No.	%
All NHL cases	1,264	100.0	700	100.0
Diffuse large B-cell lymphoma	215	17.0	127	18.1
Follicular lymphoma	162	12.8	72	10.3
CLL/SLL	382	30.2	218	31.1
Plasma cell neoplasms	243	19.2	144	20.6
Other NHL subtype	262	20.7	139	19.9

Abbreviations: CLL, chronic lymphocytic leukemia; NHL, non-Hodgkin lymphoma; SLL, small lymphocytic lymphoma.

^aLymphoid neoplasms were classified according the World Health Organization-based system published by the International Lymphoma Epidemiology Consortium (35).

Risk was increased at older ages for NHL and the NHL subtypes and was significantly higher among males compared with females for NHL overall (hazard ratio (HR) = 1.44, 95% confidence interval (CI): 1.28, 1.61) and for all subtypes except follicular lymphoma (Table 2). Nonwhite race/ethnicity was associated with lower risk of NHL overall (HR = 0.74, 95% CI: 0.61, 0.90) and for all subtypes except plasma cell neoplasms (HR = 1.58, 95% CI: 1.13, 2.21). Socioeconomic status, as measured by education, was not associated with NHL overall or any NHL subtype. Risk estimates for these demographic characteristics did not change appreciably after exclusion of the first year of follow-up (data not shown).

Table 2.

Risk of Non-Hodgkin Lymphoma Associated With Demographic Characteristics Among142,982 Participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, 1993–2006

	Total Study Population (N = 142,982)			All NHL Cases (n = 1,264)			DLBCL (n = 215)			Follicular Lymphoma (n = 162)			CLL/SLL (n = 382)			Plasma Cell Neoplasms (n = 243)
	No.a	%	Person-Years	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI
Age, years
≤59	47,730	33.4	381,539	265	1.00	Referent	34	1.00	Referent	52	1.00	Referent	85	1.00	Referent	52	1.00	Referent
60–64	44,140	30.9	386,517	391	1.42	1.21, 1.66	76	2.11	1.40, 3.17	44	0.84	0.56, 1.25	103	1.18	0.88, 1.58	70	1.24	0.87, 1.79
65–69	32,177	22.5	274,600	361	1.84	1.57, 2.16	59	2.30	1.50, 3.52	38	1.02	0.67, 1.55	122	1.97	1.49, 2.60	77	1.92	1.34, 2.74
70–74	18,935	13.2	155,093	244	2.24	1.88, 2.67	46	3.24	2.07, 5.06	28	1.35	0.85, 2.14	71	2.06	1.50, 2.83	43	1.91	1.27, 2.87
Sex
Female	70,905	49.6	599,308	517	1.00	Referent	86	1.00	Referent	74	1.00	Referent	163	1.00	Referent	100	1.00	Referent
Male	72,077	50.4	598,440	744	1.44	1.28, 1.61	129	1.49	1.13, 1.97	88	1.17	0.86, 1.61	218	1.33	1.08, 1.64	142	1.43	1.11, 1.85
Race/ethnicity
White, non-Hispanic	126,220	88.3	1,061,788	1,151	1.00	Referent	202	1.00	Referent	156	1.00	Referent	361	1.00	Referent	201	1.00	Referent
Other/unknown	16,762	11.7	135,961	110	0.74	0.61, 0.90	13	0.50	0.28, 0.87	6	0.30	0.13, 0.68	20	0.43	0.27, 0.67	41	1.58	1.13, 2.21
Education
High school or less	43,335	30.4	359,303	398	1.00	Referent	72	1.00	Referent	49	1.00	Referent	123	1.00	Referent	83	1.00	Referent
Post-high school training (not college) or some college	48,981	34.4	409,519	403	0.91	0.79, 1.04	64	0.80	0.57, 1.12	49	0.87	0.59, 1.30	114	0.83	0.64, 1.07	78	0.86	0.63, 1.17
College graduation or postgraduate study	50,282	35.3	428,927	460	0.96	0.84, 1.10	79	0.91	0.66, 1.25	64	1.07	0.73, 1.56	144	0.97	0.76, 1.24	81	0.83	0.61, 1.13

Abbreviations: CI, confidence interval; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; HR, hazard ratio; NHL, non-Hodgkin lymphoma; SLL, small lymphocytic lymphoma.

^aNumbers of cases may not sum to totals because of missing values for some independent variables.

^bHazard ratio from a Cox proportional hazards regression model. All hazard ratios and 95% confidence intervals were adjusted for age (continuous, except where estimates for categorical age are given), sex, race/ethnicity, and education.

Anthropometric factors

Analysis of men and women combined showed increasing body mass index at baseline to be positively associated with NHL overall (P_trend < 0.01) (Table 3), with an elevated risk for overweight (body mass index 25–29.9; HR = 1.16, 95% CI: 1.02, 1.33) and obesity (body mass index ≥30; HR = 1.32, 95% CI: 1.13, 1.54) as compared with normal weight (body mass index 18.5–24.9). Weight at baseline was associated with NHL (P_trend < 0.001), with increased risk in the third (HR = 1.35, 95% CI: 1.16, 1.58) and fourth (HR = 1.40, 95% CI: 1.19, 1.65) quartiles as compared with the first, and height was positively associated with NHL risk (P_trend < 0.01). Between age 20 years and baseline, 33.8% of participants remained in the same body mass index category, 2.0% moved into a lower category, and 64.3% moved into a higher category. Lifetime weight change (Table 3 and Web Tables 2–4) was moderately associated with risk of NHL overall (gain of 4.1–6 kg every 10 years: HR = 1.30, 95% CI: 1.11, 1.54), although no linear trend was detected. Similar results were observed for weight change from age 20 years to age 50 years and from age 50 years to baseline (data not shown). In analyses by NHL subtype, risk estimates for all anthropometric measures were generally higher for diffuse large B-cell lymphoma and plasma cell neoplasms than for follicular lymphoma and CLL/SLL. Anthropometric characteristics at ages 20 and 50 years (Table 3 and Web Table 5) showed similar effects on the risk of NHL and its subtypes compared with baseline, with neither body mass index nor weight emerging as the dominant influence on risk. Analyses by sex (Web Tables 2 and 3) showed that greater height was a risk factor only in men (top quartile vs. bottom: for men, HR = 1.46, 95% CI: 1.13, 1.88; for women, HR = 1.00, 95% CI: 0.79, 1.25; P for difference = 0.04). Risk estimates for body mass index and weight at each age were also slightly higher among men than among women, but these differences were not significant (Web Tables 2 and 3).

Table 3.

Risk of Non-Hodgkin Lymphoma Associated With Anthropometric Risk Factors at Age 20 Years and Baseline Among 142,982 Participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, 1993–2006

	Person-Years of Follow-up	All NHL Cases (n = 1,264)			DLBCL (n = 215)			Follicular Lymphoma (n = 162)			CLL/SLL (n = 382)			Plasma Cell Neoplasms (n = 243)
		No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI
WHO BMIc at age 20 years
<18.5	89,607	72	0.84	0.66, 1.06	17	1.22	0.74, 2.02	10	0.85	0.45, 1.62	21	0.81	0.52, 1.26	12	0.71	0.40, 1.29
18.5–24.9	905,674	940	1.00	Referent	157	1.00	Referent	129	1.00	Referent	286	1.00	Referent	173	1.00	Referent
25–29.9	160,178	206	1.19	1.02, 1.38	35	1.19	0.82, 1.73	21	0.88	0.55, 1.41	63	1.20	0.91, 1.59	41	1.33	0.94, 1.88
≥30	21,718	23	1.09	0.72, 1.65	1	N/A	N/A	2	N/A	N/A	5	N/A	N/A	12	3.08	1.71, 5.54
Ptrendd		<0.001			0.230			0.288			0.140			<0.001
Weighte at age 20 years, kg
Quartile 1	297,979	257	1.00	Referent	40	1.00	Referent	31	1.00	Referent	78	1.00	Referent	55	1.00	Referent
Quartile 2	225,541	232	1.14	0.95, 1.36	41	1.26	0.81, 1.95	24	0.92	0.54, 1.58	65	1.01	0.73, 1.41	43	1.09	0.73, 1.63
Quartile 3	357,725	392	1.37	1.17, 1.61	66	1.46	0.98, 2.17	62	1.58	1.02, 2.45	124	1.34	1.01, 1.79	58	1.06	0.73, 1.55
Quartile 4	305,482	370	1.49	1.27, 1.76	65	1.67	1.12, 2.50	45	1.32	0.83, 2.10	112	1.41	1.06, 1.89	84	1.80	1.27, 2.55
Ptrendd		<0.001			0.013			0.182			0.033			<0.001
WHO baseline BMI
<18.5	8,295	9	1.23	0.64, 2.39	4	N/A	N/A	0	N/A	N/A	3	N/A	N/A	2	N/A	N/A
18.5–24.9	384,159	351	1.00	Referent	58	1.00	Referent	49	1.00	Referent	110	1.00	Referent	57	1.00	Referent
25–29.9	507,204	564	1.16	1.02, 1.33	87	1.07	0.76, 1.50	76	1.14	0.79, 1.65	169	1.12	0.88, 1.43	112	1.45	1.05, 2.01
≥30	279,347	321	1.32	1.13, 1.54	63	1.58	1.10, 2.27	36	1.03	0.67, 1.60	95	1.25	0.95, 1.65	66	1.69	1.18, 2.41
Ptrendd		<0.01			0.056			0.465			0.746			<0.01
Baseline weighte, kg
Quartile 1	328,108	296	1.00	Referent	51	1.00	Referent	37	1.00	Referent	89	1.00	Referent	52	1.00	Referent
Quartile 2	283,105	288	1.15	0.98, 1.36	46	1.05	0.71, 1.57	44	1.32	0.85, 2.04	90	1.16	0.86, 1.55	51	1.26	0.85, 1.85
Quartile 3	299,518	353	1.35	1.16, 1.58	54	1.18	0.81, 1.74	45	1.27	0.82, 1.97	102	1.25	0.94, 1.67	79	1.88	1.32, 2.69
Quartile 4	278,182	318	1.40	1.19, 1.65	63	1.63	1.12, 2.37	36	1.14	0.72, 1.82	98	1.39	1.04, 1.86	59	1.59	1.09, 2.32
Ptrendd		<0.001			<0.01			0.555			0.215			<0.01
Baseline heightf, cm
Quartile 1	260,243	263	1.00	Referent	38	1.00	Referent	32	1.00	Referent	84	1.00	Referent	55	1.00	Referent
Quartile 2	418,916	401	0.89	0.76, 1.04	62	0.92	0.61, 1.39	56	0.95	0.61, 1.48	116	0.77	0.58, 1.02	71	0.86	0.60, 1.24
Quartile 3	232,790	280	1.10	0.93, 1.31	53	1.40	0.91, 2.15	34	1.01	0.62, 1.66	78	0.91	0.66, 1.24	60	1.33	0.91, 1.95
Quartile 4	277,976	310	1.19	1.00, 1.40	59	1.56	1.03, 2.36	40	1.07	0.67, 1.71	102	1.14	0.85, 1.53	54	1.15	0.78, 1.69
Ptrendd		<0.01			<0.01			0.571			0.023			0.152
Change in weight per 10 yearsg
Loss	77,634	90	1.08	0.85, 1.37	10	0.70	0.35, 1.39	18	1.72	0.97, 3.05	26	1.00	0.64, 1.56	16	1.00	0.56, 1.78
Gain of 0–2 kg	301,501	305	1.00	Referent	53	1.00	Referent	40	1.00	Referent	97	1.00	Referent	52	1.00	Referent
Gain of 2.1–4 kg	357,600	387	1.14	0.98, 1.32	66	1.13	0.78, 1.63	52	1.12	0.74, 1.70	114	1.03	0.78, 1.35	78	1.40	0.98, 2.00
Gain of 4.1–6 kg	235,519	273	1.30	1.11, 1.54	46	1.32	0.88, 1.97	23	0.77	0.46, 1.30	91	1.34	1.00, 1.79	51	1.48	1.00, 2.20
Gain of >6 kg	210,479	192	1.15	0.96, 1.39	37	1.41	0.91, 2.18	29	1.16	0.71, 1.90	49	0.88	0.62, 1.26	42	1.55	1.02, 2.36
Ptrendd		0.153			0.114			0.946			0.946			0.216

Abbreviations: BMI, body mass index; CI, confidence interval; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; HR, hazard ratio; N/A, not applicable; NHL, non-Hodgkin lymphoma; SLL, small lymphocytic lymphoma; WHO, World Health Organization.

^aNumbers of cases may not sum to totals because of missing values for some independent variables.

^bHazard ratio from a Cox proportional hazards regression model. Hazard ratios were not calculated where the sample size was less than 5. Hazard ratios and 95% confidence intervals were adjusted for age (continuous), sex, race/ethnicity, and education. Change in weight per 10 years was also adjusted for BMI at age 20 years (continuous). Mutual adjustment for smoking and alcohol consumption did not alter risk estimates by more than 10%; therefore, these factors were excluded from the final models.

^cWeight (kg)/height (m)².

^dP_trend value from an adjusted test for trend performed using a Cox model and a continuous independent variable.

^eQuartiles of weight (kg) were defined separately for men and women, as follows. Weight at age 20 years: quartile 1—men, <64.2; women, <51.9; quartile 2—men, 64.2–72.7; women, 51.9–54.5; quartile 3—men, 72.8–79.5; women, 54.6–59.1; quartile 4—men, >79.5; women, >59.1. Baseline weight: quartile 1—men, <77.4; women, <61.5; quartile 2—men, 77.4–85.5; women, 61.5–70.0; quartile 3—men, 85.6–95.5; women, 70.1–80.0; quartile 4—men, >95.5; women, >80.0.

^fQuartiles of baseline height (cm) were defined separately for men and women, as follows. Quartile 1—men, <172.8; women, <157.6; quartile 2—men, 172.8–177.8; women, 157.6–162.6; quartile 3—men, 177.9–182.9; women, 162.7–167.6; quartile 4—men, >182.9; women, >167.6.

^gChange in weight per 10 years was calculated from age 20 years to baseline.

Results were similar after exclusion of the first year of follow-up (data not shown), and anthropometric characteristics were consistent in the PLCO cohort during the study period, with 69.8% of participants reporting the same World Health Organization category of body mass index upon reassessment that they reported at baseline.

Smoking

Risk of NHL overall and most NHL subtypes was not associated with cigarette smoking as measured by current smoking status, years since quitting smoking, age at starting smoking, duration, intensity (packs/day), and cumulative lifetime smoking (pack-years), and no significant trends were observed for any of these measures of smoking behavior (Table 4). Cigarette smoking was inversely associated with follicular lymphoma (ever smoking vs. never smoking: HR = 0.62, 95% CI: 0.45, 0.85), but there was no clear pattern by intensity, duration, or cumulative lifetime exposure. The inverse association between smoking and follicular lymphoma persisted in analyses stratified by sex, age category, and level of education and after controlling for body mass index, drinking status, total number of drinks per week, grams of ethanol consumed per week, and several dietary factors (data not shown).

Table 4.

Risk of Non-Hodgkin Lymphoma Associated With Smoking Behavior Among 142,982 Participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, 1993–2006

	Person-Years of Follow-up	All NHL Cases (n = 1,264)			DLBCL (n = 215)			Follicular Lymphoma (n = 162)			CLL/SLL (n = 382)			Plasma Cell Neoplasms (n = 243)
		No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI
Smoking status
Never smoker	565,061	603	1.00	Referent	102	1.00	Referent	94	1.00	Referent	174	1.00	Referent	120	1.00	Referent
Ever smoker	632,523	657	0.94	0.84, 1.05	113	0.96	0.73, 1.26	67	0.62	0.45, 0.85	207	1.05	0.85, 1.29	122	0.87	0.67, 1.13
Current smoker	117,745	105	0.90	0.73, 1.11	13	0.68	0.38, 1.23	14	0.74	0.42, 1.31	39	1.20	0.84, 1.71	14	0.58	0.33, 1.01
Former smoker	514,778	552	0.95	0.84, 1.06	100	1.01	0.76, 1.34	53	0.59	0.42, 0.83	168	1.02	0.82, 1.27	108	0.93	0.71, 1.21
Years since quitting smoking (former smokers)
1–10	126,203	116	0.88	0.72, 1.08	23	1.06	0.67, 1.67	12	0.57	0.31, 1.05	33	0.89	0.61, 1.29	21	0.78	0.49, 1.25
11–20	130,832	136	0.95	0.79, 1.15	18	0.76	0.46, 1.25	18	0.81	0.48, 1.34	43	1.06	0.75, 1.48	29	1.01	0.67, 1.52
21–30	130,146	147	1.00	0.83, 1.20	25	1.02	0.65, 1.58	13	0.57	0.32, 1.02	47	1.12	0.81, 1.55	28	0.96	0.63, 1.45
>30	117,283	148	1.01	0.84, 1.21	34	1.35	0.91, 2.01	9	0.42	0.21, 0.84	43	1.01	0.72, 1.42	28	0.98	0.65, 1.50
Ptrendc		0.338			0.154			0.374			0.604			0.637
Age at starting smoking, years
Never smoker	565,061	603	1.00	Referent	102	1.00	Referent	94	1.00	Referent	174	1.00	Referent	120	1.00	Referent
1–16	203,795	213	0.95	0.81, 1.12	34	0.90	0.60, 1.35	24	0.67	0.42, 1.07	64	1.02	0.76, 1.37	45	1.01	0.71, 1.44
17–18	181,612	205	1.01	0.86, 1.19	39	1.14	0.78, 1.66	15	0.47	0.27, 0.82	74	1.29	0.98, 1.71	36	0.90	0.62, 1.32
19–20	113,558	109	0.88	0.71, 1.08	17	0.81	0.49, 1.36	11	0.57	0.31, 1.07	30	0.85	0.58, 1.26	20	0.81	0.50, 1.30
>20	128,806	126	0.89	0.73, 1.07	23	0.96	0.61, 1.51	17	0.80	0.47, 1.34	37	0.92	0.65, 1.32	21	0.72	0.45, 1.15
Ptrendc		0.235			0.385			0.318			0.297			<0.01
Years of having smoked
Never smoker	565,061	603	1.00	Referent	102	1.00	Referent	94	1.00	Referent	174	1.00	Referent	120	1.00	Referent
1–15	151,253	152	0.94	0.79, 1.13	34	1.27	0.86, 1.88	13	0.50	0.28, 0.90	41	0.89	0.63, 1.26	29	0.92	0.61, 1.38
16–30	196,641	215	0.97	0.83, 1.14	39	1.05	0.72, 1.52	23	0.68	0.43, 1.07	70	1.12	0.85, 1.49	41	0.93	0.65, 1.34
31–40	149,646	152	0.97	0.81, 1.17	20	0.77	0.48, 1.25	19	0.77	0.46, 1.26	48	1.09	0.79, 1.50	23	0.73	0.46, 1.15
>40	121,355	129	0.88	0.72, 1.06	20	0.79	0.49, 1.29	11	0.52	0.28, 0.97	44	1.07	0.76, 1.50	27	0.90	0.59, 1.38
Ptrendc		0.493			0.026			0.365			0.846			0.730
Packs of cigarettes smoked per day
Never smoker	565,061	603	1.00	Referent	102	1.00	Referent	94	1.00	Referent	174	1.00	Referent	120	1.00	Referent
0.25	163,413	156	0.93	0.78, 1.11	27	0.98	0.64, 1.50	14	0.55	0.32, 0.97	51	1.09	0.80, 1.49	32	0.90	0.61, 1.33
0.75	230,910	229	0.88	0.76, 1.03	36	0.82	0.56, 1.21	20	0.50	0.31, 0.82	81	1.11	0.85, 1.45	42	0.81	0.56, 1.15
1.25	125,391	144	1.02	0.84, 1.22	29	1.21	0.79, 1.85	21	0.94	0.58, 1.52	41	1.02	0.72, 1.44	23	0.83	0.53, 1.32
≥1.75	111,386	128	0.99	0.81, 1.21	21	0.96	0.59, 1.55	12	0.59	0.32, 1.09	34	0.93	0.64, 1.35	25	1.00	0.64, 1.56
Ptrendc		0.420			0.964			0.627			0.472			0.741
Pack-years of smoking
Never smoker	565,061	603	1.00	Referent	102	1.00	Referent	94	1.00	Referent	174	1.00	Referent	120	1.00	Referent
<8.75	161,046	152	0.91	0.76, 1.09	25	0.91	0.58, 1.41	13	0.50	0.28, 0.89	49	1.04	0.76, 1.43	28	0.82	0.55, 1.25
8.75–22.4	156,415	173	0.99	0.83, 1.17	38	1.29	0.88, 1.88	17	0.64	0.38, 1.07	53	1.07	0.78, 1.46	39	1.10	0.76, 1.59
22.5–39.75	151,920	158	0.94	0.79, 1.12	25	0.88	0.57, 1.37	19	0.73	0.44, 1.21	57	1.19	0.88, 1.62	22	0.65	0.41, 1.03
>39.75	148,440	165	0.95	0.80, 1.14	25	0.85	0.54, 1.33	17	0.63	0.37, 1.07	44	0.89	0.64, 1.25	31	0.93	0.62, 1.40
Ptrendc		0.891			0.169			0.949			0.565			0.619

Abbreviations: CI, confidence interval; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; HR, hazard ratio; NHL, non-Hodgkin lymphoma; SLL, small lymphocytic lymphoma.

^aNumbers of cases may not sum to totals because of missing values for some independent variables.

^bHazard ratio from a Cox proportional hazards regression model. All hazard ratios and 95% confidence intervals were adjusted for age (continuous), sex, race/ethnicity, and education. Mutual adjustment for anthropometric characteristics and alcohol consumption did not alter risk estimates by more than 10%; therefore, these factors were excluded from the final models.

^cP_trend value from an adjusted test for trend (among smokers only) performed using a Cox model and a continuous independent variable.

Results were similar after exclusion of the first year of follow-up (data not shown), and smoking behavior remained consistent in the PLCO cohort over time, with 92.7% of participants reporting the same smoking status upon reassessment that they reported at baseline.

Alcohol consumption

Among subjects who completed the dietary history questionnaire, 700 NHL cases were diagnosed during 596,280 person-years of follow-up (median, 6.4 years per person) (Table 5). Alcohol consumption was not significantly associated with the risk of NHL overall in the PLCO cohort as measured by total number of drinks per week or grams of ethanol consumed per week. Analysis of NHL subtypes showed a significant inverse association between alcohol consumption and diffuse large B-cell lymphoma as measured by total number of drinks per week (P_trend = 0.016) and grams of ethanol consumed per week (P_trend = 0.017). However, risk of diffuse large B-cell lymphoma was not elevated among nondrinkers in comparison with light drinkers, and estimates were based on a small number of cases. Risk estimates for NHL overall and NHL subtypes did not vary according to alcoholic beverage type or sex. Results were similar after controlling for smoking status and dietary risk factors and after excluding the first year of follow-up (data not shown).

Table 5.

Risk of Non-Hodgkin Lymphoma Associated With Alcohol Consumption Among 142,982 Participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, 1993–2006

	Person-Years of Follow-up	All NHL Cases (n = 700)			DLBCL (n = 127)			Follicular Lymphoma (n = 72)			CLL/SLL (n = 218)			Plasma Cell Neoplasms (n = 144)
		No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI	No.a	HRb	95% CI
Total alcohol consumption, drinks/week
Nondrinker	197,180	228	0.96	0.77, 1.19	41	0.78	0.47, 1.27	22	0.83	0.42, 1.62	74	1.21	0.79, 1.84	44	0.77	0.48, 1.23
<1	103,394	119	1.00	Referent	26	1.00	Referent	14	1.00	Referent	31	1.00	Referent	28	1.00	Referent
1–3	124,768	161	1.08	0.85, 1.37	28	0.84	0.49, 1.43	21	1.16	0.59, 2.30	51	1.30	0.83, 2.04	32	0.96	0.58, 1.61
4–13	107,635	121	0.88	0.68, 1.14	25	0.79	0.45, 1.39	8	0.48	0.20, 1.17	37	1.03	0.64, 1.68	27	0.92	0.54, 1.57
≥14	62,167	68	0.84	0.62, 1.14	7	0.37	0.16, 0.87	7	0.71	0.28, 1.79	24	1.13	0.66, 1.95	12	0.71	0.36, 1.41
Ptrendc		0.187			0.016			0.302			0.272			0.068
Beer consumption, drinks/week
Nondrinker	197,180	228	0.96	0.81, 1.14	41	0.86	0.58, 1.27	22	0.92	0.54, 1.58	74	1.07	0.79, 1.46	44	0.77	0.53, 1.13
<1	276,440	323	1.00	Referent	65	1.00	Referent	35	1.00	Referent	96	1.00	Referent	73	1.00	Referent
1–2	55,750	69	0.94	0.72, 1.22	11	0.69	0.36, 1.33	7	0.90	0.39, 2.07	22	1.00	0.62, 1.61	11	0.72	0.38, 1.37
≥3	65,774	77	0.89	0.69, 1.15	10	0.54	0.28, 1.07	8	0.90	0.40, 1.99	25	0.98	0.62, 1.55	15	0.81	0.46, 1.44
Ptrendc		0.247			0.077			0.553			0.462			0.441
Wine consumption, drinks/week
Nondrinker	197,180	228	0.99	0.83, 1.18	41	1.00	0.66, 1.53	22	0.88	0.51, 1.54	74	1.05	0.77, 1.44	44	0.81	0.54, 1.20
<1	234,031	279	1.00	Referent	50	1.00	Referent	31	1.00	Referent	87	1.00	Referent	61	1.00	Referent
1–2	76,335	101	1.14	0.91, 1.44	23	1.47	0.89, 2.42	7	0.67	0.29, 1.52	26	0.95	0.61, 1.47	23	1.20	0.74, 1.94
≥3	87,599	89	0.88	0.69, 1.12	13	0.72	0.39, 1.34	12	0.96	0.49, 1.90	30	0.95	0.62, 1.45	15	0.70	0.39, 1.24
Ptrendc		0.304			0.232			0.513			0.693			0.138
Liquor consumption, drinks/week
Nondrinker	197,180	228	0.98	0.83, 1.16	41	0.90	0.61, 1.34	22	0.90	0.53, 1.52	74	1.06	0.78, 1.42	44	0.88	0.60, 1.28
<1	306,107	355	1.00	Referent	69	1.00	Referent	40	1.00	Referent	109	1.00	Referent	71	1.00	Referent
1–2	33,698	46	1.09	0.80, 1.49	6	0.70	0.30, 1.62	4	N/A	N/A	17	1.29	0.77, 2.15	10	1.28	0.66, 2.50
≥3	58,160	68	0.89	0.69, 1.16	11	0.70	0.37, 1.32	6	0.71	0.30, 1.69	17	0.71	0.43, 1.19	18	1.29	0.77, 2.18
Ptrendc		0.455			0.219			0.395			0.058			<0.01
Total ethanol consumption, g/week
Nondrinker	197,180	228	0.96	0.77, 1.21	41	0.77	0.47, 1.26	22	0.79	0.40, 1.54	74	1.23	0.80, 1.90	44	0.79	0.48, 1.28
2.9–11.8	98,911	113	1.00	Referent	25	1.00	Referent	14	1.00	Referent	29	1.00	Referent	26	1.00	Referent
11.9–37.3	100,947	133	1.11	0.86, 1.43	27	1.00	0.58, 1.72	14	0.92	0.44, 1.94	41	1.32	0.82, 2.13	26	1.00	0.58, 1.73
37.4–123.9	99,109	105	0.86	0.66, 1.12	14	0.50	0.26, 0.96	13	0.84	0.39, 1.80	36	1.14	0.70, 1.87	22	0.85	0.48, 1.51
>123.9	98,999	118	0.92	0.71, 1.20	20	0.66	0.36, 1.21	9	0.55	0.24, 1.30	37	1.12	0.68, 1.84	25	0.96	0.55, 1.68
Ptrendc		0.199			0.017			0.295			0.239			0.043

Abbreviations: CI, confidence interval; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; HR, hazard ratio; N/A, not applicable; NHL, non-Hodgkin lymphoma; SLL, small lymphocytic lymphoma.

^aNumbers of cases may not sum to totals because of missing values for some independent variables.

^bHazard ratio from a Cox proportional hazards regression model. All hazard ratios and 95% confidence intervals were adjusted for age (continuous), sex, race/ethnicity, and education. Hazard ratios were not calculated where the sample size was less than 5. Mutual adjustment for anthropometric characteristics and smoking did not alter risk estimates by more than 10%; therefore, these factors were excluded from the final models.

^cP_trend value from an adjusted test for trend performed using a Cox model and a continuous independent variable. P_trend was calculated among drinkers only for total drinks per week and among only drinkers of each type of beverage for beer, wine, and liquor.

DISCUSSION

In this large, population-based prospective cohort study of older adults, enlarged body size, as measured by body mass index, height, and weight, was associated with elevated risk of NHL, with height being a risk factor only in men. Smoking was not related to NHL overall but was inversely associated with follicular lymphoma, and alcohol consumption was unrelated to NHL risk.

Our study adds to existing evidence suggesting that increased body mass index is modestly associated with elevated NHL risk (9–18). Our observation of increased risk of NHL with greater height also agrees with prior reports (13, 14, 18, 23, 28, 36). Anthropometry-related risk of NHL is not consistent in the literature, however. Several prior studies have shown no association of body mass index (19–23) or height (12, 19) with NHL, and to our knowledge only 1 other study has shown increased risk of NHL associated with higher weight (18). Results of our sex-specific analyses of body mass index and NHL risk were consistent with those of prior studies, showing slightly higher risk among men than among women (17), although our male-specific height finding contradicts the strong height-related risk detected in a large all-female cohort (18). Our nonsignificant findings for lifetime weight change were similar to those reported previously (14, 23). Results of our analyses by NHL subtype were consistent with those of prior studies, with somewhat higher risks for diffuse large B-cell lymphoma and plasma cell neoplasms as compared with CLL/SLL and follicular lymphoma (12–15, 21).

Several plausible mechanisms exist through which enlarged body size might increase NHL risk. Greater height may be an indicator of childhood nutrition patterns that could indirectly affect NHL risk—for example, by altering the likelihood of childhood infection (37). Alternatively, elevated body mass index and weight may be markers for certain dietary patterns, insulin resistance, inflammation, and lower levels of physical activity that may influence cancer risk (38). Future studies of anthropometric factors and NHL would benefit from assessing these parameters in conjunction with weight and height.

Our observation that smoking was unrelated to the risk of NHL overall is consistent with prior case-control (7) and cohort (24–28) studies. The inverse association we noted between smoking and follicular lymphoma was unexpected, however. Case-control (7) and some cohort (24, 25) studies have shown smoking to be associated with an increased risk of follicular lymphoma, although at least 1 cohort study showed no association between smoking and follicular lymphoma (26) and 1 found an inverse association (28) similar to our observation in the PLCO Trial. Our failure to detect a dose-response with intensity, duration, or cumulative lifetime exposure to cigarette smoking suggests confounding rather than a true protective association. However, the inverse association we observed between smoking and follicular lymphoma persisted after we accounted for alcohol, body mass index, and dietary factors that might lower the risk of NHL or its subtypes. Other sources of confounding that we were unable to control for may include immunity-modulating comorbid conditions, such as atopy, that have been shown to be protective against follicular lymphoma (39). Although it is likely that the inverse association between follicular lymphoma and smoking was due to chance or unmeasured confounding, the subtype specificity and the consistency with another study of older US adults (28) suggests that additional research in prospective studies is warranted.

We did not observe any significant association between total alcohol consumption and risk of NHL overall in the PLCO cohort, nor did we observe any association between beverage type and NHL risk, even after controlling for smoking status and dietary factors. Because of the unusually large number of nondrinkers in the PLCO cohort, we were able to separate nondrinkers (<1 drink/month) from light drinkers (<1 drink/week). Under the hypothesis that alcohol is inversely associated with NHL risk, we expected to observe increased risk among nondrinkers as compared with light drinkers and progressively decreasing risk as consumption increased. Although our risk estimates were generally less than 1 for participants who consumed the highest amounts of alcohol, we did not observe any significant associations for NHL overall, and the inverse association we observed for diffuse large B-cell lymphoma was based on few cases. Our overall null results cannot rule out the hypothesis that the inverse association observed in case-control studies (8) is part of a prodrome in which patients reduce or eliminate alcohol consumption prior to diagnosis. However, several other large cohorts with similar levels of alcohol consumption have shown an inverse association between alcohol and NHL risk (28–30). In addition, high levels of alcohol consumption (≥300 g/week) were associated with significantly decreased risk of NHL in a Japanese cohort (31). However, similar to our observation in the PLCO Trial, risk among nondrinkers in the Japanese cohort was not elevated relative to light drinkers. Decreased risk of NHL associated with alcohol consumption is supported by evidence that alcohol modulates the immune system (40) and retards the growth of malignant lymphoid cells via inhibition of the mammalian target of rapamycin (mTOR) (41). In light of such evidence, in future studies of alcohol use and NHL risk, investigators should combine data on drinking behavior with biologic measures of exposure that may affect alcohol-related NHL risk.

The strengths of our study include its prospective design; the availability of a large number of histologically confirmed incident cases, which enabled us to analyze risks by NHL subtype; data on anthropometric factors at different ages; our ability to examine nondrinkers versus light drinkers; the length of follow-up; the consistency of anthropometric and smoking exposures in the PLCO cohort over time; and the similarity of NHL risk patterns in the PLCO cohort with the broader population of white, non-Hispanic US adults. However, our failure to observe a dominant effect of body mass index, height, or weight may have been due to underestimation of body size in self-reports, which could have attenuated risk estimates (42). Our findings of decreased risk of follicular lymphoma among smokers and no association between NHL and alcohol consumption may have resulted from residual confounding. In addition, we performed many statistical tests, and thus some findings could have been due to chance; and we did not have data on some potential confounders such as physical activity.

In summary, anthropometric factors were modestly associated with increased risk of NHL in the PLCO cohort and were slightly stronger predictors of risk in men than in women; no clear association was observed between smoking and NHL; and alcohol consumption appeared unrelated to NHL. Our data support previous studies suggesting that body mass index is associated with NHL risk. Future studies of anthropometric factors and NHL that incorporate measurement of biomarkers are essential to help elucidate potential causal mechanisms. The protective effect of smoking on follicular lymphoma has been observed previously but may have been due to residual confounding, so additional prospective studies are warranted to clarify the association between smoking and follicular lymphoma. Finally, although our findings for alcohol consumption argue against a biologic association with NHL risk, future laboratory research into the action of alcohol on the immune system may identify genetic or other biologic risk factors that could be measured in epidemiologic investigations of alcohol use and NHL risk.

Glossary

Abbreviations

CI

confidence interval

CLL

chronic lymphocytic leukemia

HR

hazard ratio

NHL

non-Hodgkin lymphoma

PLCO

Prostate, Lung, Colorectal, and Ovarian

SLL

small lymphocytic lymphoma