Smoking, alcohol use, obesity, and overall survival from non-Hodgkin lymphoma

A population-based study

Authors


Abstract

BACKGROUND:

Smoking, alcohol use, and obesity appear to increase the risk of developing non-Hodgkin lymphoma (NHL), but to the authors' knowledge, few studies to date have assessed their impact on NHL prognosis.

METHODS:

The association between prediagnosis cigarette smoking, alcohol use, and body mass index (BMI) and overall survival was evaluated in 1286 patients enrolled through population-based registries in the United States from 1998 through 2000. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were estimated using Cox regression, adjusting for clinical and demographic factors.

RESULTS:

Through 2007, 442 patients had died (34%), and the median follow-up for surviving patients was 7.7 years. Compared with never smokers, former (HR, 1.59; 95% CI, 1.12-2.26) and current (HR, 1.50; 95% CI, 0.97-2.29) smokers had poorer survival, and poorer survival was found to be positively associated with smoking duration, number of cigarettes smoked per day, pack-years of smoking, and shorter time since quitting (all P <0.01). Alcohol use was associated with poorer survival (P = 0.03); compared with nonusers. Those drinking >43.1 g/week (median intake among drinkers) had poorer survival (HR, 1.55; 95% CI, 1.06-2.27), whereas those drinkers consuming less than this amount demonstrated no survival disadvantage (HR, 1.13; 95% CI, 0.75-1.71). Greater BMI was associated with poorer survival (P = 0.046), but the survival disadvantage was only noted among obese individuals (HR, 1.32 for BMI ≥30 vs BMI 20-24.9; 95% CI, 1.02-1.70). These results held for lymphoma-specific survival and were broadly similar for diffuse large B-cell lymphoma and follicular lymphoma.

CONCLUSIONS:

NHL patients who smoked, consumed alcohol, or were obese before diagnosis were found to have a poorer overall and lymphoma-specific survival. Cancer 2010. © 2010 American Cancer Society.

In 2008, the American Cancer Society estimates that 66,120 persons will have been diagnosed with non-Hodgkin lymphoma (NHL), and 19,160 will have died from this disease.1 NHL is now the fifth most common cause of cancer among both men and women. NHL incidence rates have been increasing over much of the 20th century and that growth has only recently begun to slow. Five-year relative survival for the time period of 1974 to1995 was stable at approximately 50%, but was reported to have increased to 66% for the time period 1996 to 2004.2 Of the estimated 10.7 million cancer survivors in the United States in 2004, approximately 431,000 had NHL,2 making this an important group from a clinical and public health perspective. Diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma are the 2 most common NHL subtypes.

The strongest and most robust predictors of outcome in NHL are age, stage, number of lymph node or extranodal sites, performance status, and certain biochemical measurements (eg, serum lactate dehydrogenase and hemoglobin level), and these factors have been aggregated into various clinical prognostic indices, including the International Prognostic Index (IPI) for DLBCL3 and the Follicular IPI (FLIPI) for follicular lymphoma.4 Multiple tumor biomarkers have also been widely evaluated as predictors of NHL prognosis, although to our knowledge few are used in routine clinical practice.5, 6 More recently, host genetic background has also been evaluated, including pharmacogenetic markers,7-9 genetic markers of general immune function,10, 11 or genetic markers of DNA repair function.12 In contrast, to the best of our knowledge there are relatively few data regarding the impact of lifestyle factors on NHL prognosis. Results from large International Lymphoma Epidemiology Consortium (InterLymph) pooling projects have reported that cigarette smoking appears to slightly increase the risk of NHL overall and follicular lymphoma in particular,13 alcohol decreases the risk of NHL across most subtypes,14 and obesity increases the risk of DLBCL.15 However, to our knowledge only 3 studies to date have evaluated the role of these factors in NHL prognosis and suggest that cigarette smoking,16, 17 alcohol use,16, 17 and obesity18 may be adverse prognostic factors. Therefore, we evaluated the impact of these factors on overall and lymphoma-specific survival in a series of patients who participated in a population-based case-control study in 4 regions of the United States.

MATERIALS AND METHODS

Study Population

This study of NHL survival has been previously described.10 Briefly, subjects with newly diagnosed, histologically confirmed NHL were enrolled in a population-based case-control study from July 1998 through June 200019; the analyses presented herein focus on the cases enrolled in that study. The cases were rapidly reported from 4 Surveillance, Epidemiology, and End Results (SEER) cancer registries in the Detroit metropolitan area, the state of Iowa, Los Angeles County, and northwestern Washington State. All consecutive NHL patients diagnosed in Iowa and the Seattle area were included in this study, whereas in Los Angeles and Detroit, all African-American NHL patients were included, but only a random sample of non-African-American patients. Eligible patients had a first primary diagnosis of NHL between the ages of 20 and 74 years (inclusive), and were alive and competent to participate in the interview. Any patients known to be positive for the human immunodeficiency virus (HIV) were excluded. Local Internal Review Board approval was obtained for each of the participating recruitment centers, and consent was obtained for all participants.

Of 2248 eligible patients, 320 (14%) died before we could conduct an interview, 127 (6%) could not be located, 16 (1%) had moved out of the area, and 57 (3%) had physician refusals. We attempted to contact the remaining 1728, but 274 (16%) patients declined to be interviewed, and 133 (8%) never responded or were not interviewed because of illness, impairment, or other reasons. This left 1321 eligible patients in the study, for a participation rate of 76% of the patients we attempted to contact and an overall response rate of 59% of the living and deceased patients presumed to be eligible. Response rates were higher for patients with follicular lymphoma (67%) than DLBCL (51%). Approximately 60% of patients were interviewed within 6 months of their diagnosis, and 84% were interviewed within 1 year.

Lifestyle Data

In-home interviews were conducted using a computer-assisted personal interview (CAPI). To accommodate a large number of questions, we used a split-sample design, with a core set of questions given to all respondents and the remainder given to participants in either Group A (all African-American and 50% of non-African-American participants) or Group B (50% of non-African-American participants). Before the in-person interview, participants were mailed a form for listing residential and job history, and either a family medical history questionnaire (Group A) or a diet and lifestyle questionnaire (Group B). During the interview, the interviewer administered a CAPI that included demographics, height and weight, occupational history, pesticide exposure, and hair dye use. The Group A CAPI included an extended medical history and use of illicit drugs, whereas the Group B CAPI included an abbreviated medical history, cell phone use, and sun exposure.

Self-reported height and weight 1 year before diagnosis was obtained from all participants as part of the CAPI. Alcohol and cigarette smoking (but not other tobacco use) was only collected for Group B participants. Usual intake of beer, white wine, red wine, and liquor 1 year before diagnosis (and excluding recent changes) were obtained as part of the self-administered modified Block 1995 Health Habits and History Questionnaire.20 Total grams of alcohol intake per week were estimated from the Block database. The smoking history included ever smoking, age at start, age at stop, and average number of cigarettes smoked per day.

Clinical and Outcome Data

Date of diagnosis, histology, stage, presence of B-symptoms, first course of therapy, date of last follow-up, and vital status were derived from linkage to registry databases at each study site. Histology was coded initially according to the International Classification of Diseases-Oncology (ICD-O), second edition,21 and this was later updated to the third edition by each registry.22 We grouped cases into NHL subtypes according to the InterLymph guidelines.23 Data regarding the first course of therapy included use of single-agent or multi-agent chemotherapy, radiation, other therapies exclusive of chemotherapy and/or radiation, and no therapy (presumed to be observation); information regarding individual agents and doses was not available. The SEER registries collect date and cause of death, but do not collect data concerning treatment response or disease recurrence or progression. In 2007, we conducted a second linkage to update survival information. Of the 1321 cases in the original study, we had clinical and follow-up data available for 1286 cases for this analysis.

Statistical Analysis

Body mass index (BMI) was calculated as weight (kg) divided by height (m) squared, and was categorized a priori according to World Health Organization categories of <20, 20 to 24.9, 25.0 to 29.9, and 30+ kg/m2. Alcohol use was categorized as nondrinker, alcohol use ≤median of all alcohol users in this study (43.1 g/week), or alcohol use >median of alcohol users; in secondary analyses, we also modeled alcohol as a continuous variable. Smoking was categorized by status (never, former, or current), duration (never smoker, <20 years, or ≥20 years), intensity (never smoker, ≤20 [1 pack] cigarettes/day, or ≥21 cigarettes/day), pack-years (never smoker, <17 pack-years, or ≥17 pack-years), and time since quitting (never smoker, quit >20 years, quit 10-19 years, quit <10 years, or current smokers).

Overall survival was defined as the time from diagnosis to the date of death or last follow-up; patients alive at the time of last follow-up were censored at that time point. Lymphoma-specific survival was defined as death due to lymphoma; other deaths were censored at date of death. We used Cox proportional hazards regression24 to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) after adjusting for clinical and demographic variables (see Table 1 for variables and categories). These analyses were conducted for the entire cohort as well as the subset groups of DLBCL and follicular lymphoma.

Table 1. Univariate Associations of Demographic and Clinical Factors, NCI-SEER NHL Survival Study, 1998-2007
FactorAll NHLDLBCLFollicular Lymphoma
No.% DeadHR (95% CI)No.% DeadHR (95% CI)No.% DeadHR (95% CI)
  1. NCI-SEER indicates National Cancer Institute-Surveillance, Epidemiology, and End Results; NHL, non-Hodgkin lymphoma; DLBCL, diffuse large B-cell lymphoma; HR, hazard ratio; 95% CI, 95% confidence interval; CLL, chronic lymphocytic leukemia; SLL, small lymphocytic lymphoma.

Age, y
 ≤6071425.4%1.00 (reference)24723.5%1.00 (reference)19221.4%1.00 (reference)
 >6057245.6%1.98 (1.63-2.39)17349.7%2.41 (1.72-3.36)13638.2%1.86 (1.23-2.80)
Sex
 Female59531.9%1.00 (reference)18933.3%1.00 (reference)16529.7%1.00 (reference)
 Male69136.5%1.17 (0.97-1.41)23135.1%1.03 (0.74-1.43)16327.0%0.93 (0.62-1.39)
Race
 White109733.5%1.00 (reference)36833.7%1.00 (reference)28727.5%1.00 (reference)
 Nonwhite18939.7%1.24 (0.97-1.59)5238.5%1.20 (0.75-1.92)4134.1%1.33 (0.76-2.36)
Education level, y
 <1212749.6%1.00 (reference)4553.3%1.00 (reference)2532.0%1.00 (reference)
 12-1579633.2%0.58 (0.44-0.77)25230.2%0.45 (0.29-0.72)21328.6%0.89 (0.43-1.87)
 ≥1636231.8%0.57 (0.42-0.77)12236.1%0.59 (0.36-0.96)9026.7%0.83 (0.37-1.84)
B symptoms
 No49629.6%1.00 (reference)14628.8%1.00 (reference)16025.6%1.00 (reference)
 Yes25441.7%1.62 (1.26-2.08)11839.0%1.54 (1.01-2.34)4729.8%1.24 (0.67-2.27)
 Unknown53635.3%1.25 (1.01-1.55)15635.9%1.34 (0.90-2.00)12136.4%1.26 (0.81-1.96)
Stage of disease
 Local/regional55626.4%1.00 (reference)24030.8%1.00 (reference)13018.5%1.00 (reference)
 Distant66342.2%1.80 (1.48-2.20)16738.9%1.39 (0.99-1.94)18135.4%2.07 (1.29-3.30)
 Unknown6722.4%0.81 (0.48-1.38)1338.5%1.27 (0.52-3.15)1729.4%1.51 (0.57-3.95)
NHL subtype
 DLBCL42034.3%1.00 (reference)      
 Follicular32828.4%0.77 (0.59-1.00)      
 Mantle cell5463.0%2.36 (1.62-3.43)      
 Marginal zone11525.2%0.65 (0.43-0.96)      
 Peripheral T-cell4240.5%1.31 (0.79-2.17)      
 CLL/SLL12850.8%1.37 (1.03-1.82)      
 Other subtypes19930.2%0.91 (0.66-1.24)      

RESULTS

There were 1286 patients with a median age of 58 years (range, 20-74 years) at diagnosis between 1998 and 2000. A slight majority of subjects were male (54%). The most common histologies were DLBCL (33%), follicular (26%), chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL) (10%), marginal zone (9%), mantle cell (4%), and peripheral T-cell (3%). As part of their initial therapy, 68% of patients received some type of chemotherapy, whereas 27% received radiotherapy (treatments not mutually exclusive).

Through 2007, 442 (34%) patients died, including 144 of 420 DLBCL (34%) and 93 of 328 follicular lymphoma (28%) patients. The median follow-up for surviving subjects was 7.7 years (range, 1.3-9.5 years). Age >60 years, male sex, nonwhite race, lower education, presence of B-symptoms, distant stage, and NHL subtype (DLBCL, mantle cell, peripheral T-cell, and CLL/SLL) were found to be associated with poorer survival on univariate analysis for all NHL patients, although not all were statistically significant at P <.05 (Table 1).

Data regarding smoking and alcohol use were only collected for the patients who were part of the Group B arm of the study (N = 550). Of the 471 patients from this arm of the study with available smoking data, 34% were former smokers and 19% were current smokers at the time of diagnosis. After adjustment for clinical and demographic factors in Table 1, former (HR, 1.59; 95% CI, 1.12-2.26) and current (HR, 1.50; 95% CI, 0.97-2.29) smokers were found to have poorer survival compared with never smokers (Table 2). Poorer survival was positively associated with longer smoking duration (P <.001), greater number of cigarettes smoked per day (P = .009), greater pack-years of smoking (P = .0005), and shorter time since quitting (P = .002). It is interesting to note that compared with never smokers, patients who had quit >20 years before diagnosis did not have a higher risk of death (HR, 1.14; 95% CI, 0.70-1.86). All these associations were somewhat weaker for DLBCL and stronger for follicular lymphoma (Table 2).

Table 2. Multivariate Adjusted HRs and 95% CIsa for the Association of Smoking, Alcohol Use, and BMI With Overall Survival for All NHL and for DLBCL and Follicular Lymphoma, NCI-SEER NHL Survival Study, 1998-2007
 All NHLDLBCLFollicular Lymphoma
No.% DeadHR (95% CI)No.% DeadHR (95% CI)No.% DeadHR (95% CI)
  • HRs indicate hazard ratios; 95% CIs, 95% confidence intervals; BMI, body mass index; NHL, non-Hodgkin lymphoma; DLBCL, diffuse large B-cell lymphoma; NCI-SEER, National Cancer Institute-Surveillance, Epidemiology, and End Results; WHO, World Health Organization.

  • a

    Adjusted for age at diagnosis (≤60 years vs >60 years), gender, race, education, study site, stage of disease (distant vs not), chemotherapy, radiation, B symptoms, and histology (for the All NHL group only).

  • b

    Median intake of users; approximately equal to 3.3 cans (355 mL) of beer, 4.6 glasses (118 mL) of wine, or 2.7 shots (44 mL) of liquor per week.

  • c

    One standard deviation (5 kg/m2) change.

Smoking and alcohol use for Group B subjects
 Smoking status
  Never smoker22030.0%1.00 (reference)8632.6%1.00 (reference)5822.4%1.00 (reference)
  Former smoker16238.3%1.59 (1.12-2.26)5440.7%1.36 (0.77-2.40)4025.0%1.98 (0.81-4.80)
  Current smoker8937.1%1.50 (0.97-2.29)3033.3%1.22 (0.58-2.58)2642.3%2.95 (1.27-6.81)
 Smoking duration
  Nonsmoker22030.0%1.00 (reference)8632.6%1.00 (reference)5822.4%1.00 (reference)
  Smoked <20 y8731.0%1.09 (0.67-1.78)2931.0%1.01 (0.47-2.17)2425.0%2.04 (0.74-5.67)
  Smoked ≥20 y15944.0%1.76 (1.25-2.47)5242.3%1.53 (0.87-2.69)4134.1%2.48 (1.11-5.52)
   P for trend = .0009  P for trend = .16  P for trend = .03
 Smoking intensity (cigarettes/d)
  Nonsmoker22030.0%1.00 (reference)8632.6%1.00 (reference)5822.4%1.00 (reference)
  Smoked ≤1pack/d15135.1%1.43 (1.00-2.06)4637.0%1.17 (0.64-2.14)4228.6%2.02 (0.88-4.61)
  Smoked >1 pack/d9940.4%1.65 (1.11-2.45)3836.8%1.39 (0.72-2.66)2231.8%2.49 (0.93-6.64)
   P for trend = .009  P for trend = .32  P for trend = .04
 Smoking, pack-y
  Nonsmoker22030.0%1.00 (reference)8632.6%1.00 (reference)5822.4%1.00 (reference)
  <17 pack-y9324.7%0.96 (0.59-1.54)2927.6%0.82 (0.37-1.82)2317.4%1.29 (0.40-4.11)
  ≥17 pack-y15044.7%1.88 (1.33-2.66)5044.0%1.68 (0.95-2.95)4141.5%2.61 (1.19-5.74)
   P for trend = .0005  P for trend = .10  P for trend = .02
Time since quitting smoking (to diagnosis)
  Never smoker22030.0%1.00 (reference)8632.6%1.00 (reference)5822.4%1.00 (reference)
  Quit >20 y7031.4%1.14 (0.70-1.86)2142.9%1.16 (0.54-2.49)1717.6%1.19 (0.33-4.27)
  Quit 10-19 y5137.3%1.41 (0.85-2.36)2142.9%1.41 (0.66-3.00)1020.0%1.79 (0.38-8.40)
  Quit <10 y3951.3%3.56 (2.12-5.97)1136.4%2.24 (0.76-6.62)1233.3%3.73 (1.07-13.0)
  Current smoker8937.1%1.51 (0.99-2.32)3033.3%1.22 (0.58-2.59)2642.3%3.01 (1.30-7.00)
   P for trend = .002  P for trend = .312  P for trend = .005
 Alcohol use
  None23433.3%1.00 (reference)8337.3%1.00 (reference)6628.8%1.00 (reference)
  ≤43.1 g/wkb11230.4%1.13 (0.75-1.71)4533.3%1.24 (0.65-2.35)3010.0%0.35 (0.10-1.20)
  >43.1 g/wk11239.3%1.55 (1.06-2.27)3834.2%1.48 (0.75-2.94)2642.3%2.16 (0.98-4.77)
   P for trend = .03  P for trend = .25  P for trend = .16
BMI for Group A and B subjects combined
 BMI continuousc118933.9%1.09 (1.00-1.20)39133.8%1.07 (0.93-1.23)29931.1%1.14 (0.92-1.41)
 BMI WHO categories:
  Underweight (<20 kg/m2)5529.1%1.00 (0.59-1.68)1723.5%1.20 (0.42-3.42)1526.7%0.76 (0.26-2.23)
  Normal (20-24.9 kg/m2)36831.3%1.00 (reference)10830.6%1.00 (reference)10723.4%1.00 (reference)
  Overweight (25-29.9 kg/m2)46233.8%1.03 (0.81-1.31)14332.2%1.05 (0.67-1.64)11430.7%1.23 (073-2.08)
  Obese (≥30 kg/m2)30438.2%1.32 (1.02-1.70)12339.8%1.37 (0.88-2.14)6330.2%1.49 (0.81-2.72)
   P for trend = .046  P for trend = .20  P for trend = .11
BMI for Group B subjects only
 BMI continuousc45933.1%1.08 (0.93-1.25)16735.3%1.05 (0.85-1.30)12226.2%1.05 (0.77-1.44)
 BMI WHO categories:
  Underweight (<20 kg/m2)2128.6%1.01 (0.43-2.36)540.0%1.87 (0.42-8.22)812.5%0.38 (0.05-3.04)
  Normal (20-24.9 kg/m2)15832.3%1.00 (reference)5232.7%1.00 (reference)4827.1%1.00 (reference)
  Overweight (25-29.9 kg/m2)18531.4%0.88 (0.60-1.28)6434.4%0.93 (0.49-1.77)4623.9%1.01 (0.44-2.30)
  Obese (≥30 kg/m2)9538.9%1.33 (0.86-2.03)4639.1%1.24 (0.63-2.46)2035.0%1.29 (0.51-3.27)
   P for trend = .29  P for trend = .78  P for trend = .33

Of the 458 patients with available data regarding alcohol use from the Group B arm of the study, 49% consumed alcohol 1 year before diagnosis, and among alcohol users, the median intake was 43.1 g/week. After multivariable adjustment for demographic and clinical factors, there was poorer survival noted for drinkers of >43.1 g/week compared with never drinkers (HR, 1.55; 95% CI, 1.06-2.27), whereas there was no association for drinkers of ≤43.1 g/week (HR, 1.13; 95% CI, 0.75-1.71) (Table 2). These trends were evident for DLBCL and follicular lymphoma (Table 2), although the point estimates and trend tests did not reach conventional levels of statistical significance at P <.05. When we modeled alcohol as a continuous variable, the P values were .038 for all NHL, .17 for DLBCL, and .008 for follicular lymphoma. When we simultaneously modeled wine, beer, and liquor use, no 1 specific type of alcohol was found to be associated with NHL survival or DLBCL survival (data not shown); however, for follicular survival, there appeared to be a specific association with liquor use (HR, 2.90; 95% CI, 1.19-7.07). However, due to relatively small numbers, these results should be interpreted with caution.

All 1286 patients were asked to provide their height and weight 1 year before diagnosis, and 1189 had useable data. Based on BMI, 5% of patients were classified as underweight, 31% as being of normal weight, 39% as overweight, and 26% as obese. As shown in Table 2, after adjustment for clinical and demographic factors, only obese patients demonstrated a survival disadvantage (HR, 1.32 compared with normal weight; 95% CI, 1.02-1.70). These associations were similar for the patients only in the Group B arm (who had smoking and alcohol data), and were also similar for DLBCL and follicular lymphoma patients (Table 2). There were no associations noted with height or weight for NHL overall or by subtype (all P trend >.15; data not shown).

We next analyzed the associations in Table 2 based instead on death due to lymphoma (N = 273), and the patterns of association were found to be very similar, with the major exceptions that the association between smoking and survival were strengthened for follicular lymphoma and the association between BMI and survival strengthened for all outcomes (data not shown).

When we included pack-years of smoking, alcohol use, and BMI in the same model along with the clinical and demographic variables, the HR remained unchanged for smoking (HR, 1.87 for ≥17 pack-years vs never smoker; 95% CI, 1.29-2.72), nearly identical for BMI (HR, 1.38 for BMI >30 vs BMI of 20-24.9 kg/m2; 95% CI, 0.89-2.15), and slightly attenuated for alcohol use (HR, 1.41 for >43.1 g/week vs no alcohol; 95% CI, 0.93-2.15). Similar results were observed for DLBCL and follicular lymphoma (data not shown). These results provide some evidence that these associations were independent of each other, although a much larger sample size would be required to provide a more definitive assessment with respect to statistical significance.

DISCUSSION

In this analysis of a large, population-based sample of NHL cases, we found that smoking cigarettes, drinking >43.1 g of alcohol per week (approximately 3.3 cans [355 mL] of beer, 4.6 glasses [118 mL] of wine, or 2.7 shots [44 mL] of liquor per week), and being obese before diagnosis each had a negative impact on overall survival after accounting for clinical and demographic variables. Associations remained after the simultaneous inclusion of all 3 variables in the same statistical model, although some of the associations were no longer statistically significant, and therefore the latter observation is only preliminary. Finally, these associations held for lymphoma-specific survival, and were found to be broadly similar for both DLBCL and follicular lymphoma, with the exception that the associations with smoking were stronger for follicular lymphoma.

To our knowledge, the current study is the first report to simultaneously assess these associations on survival and to evaluate them in a North American population; replication is needed. Although based on observational data, there do not appear to be obvious explanations for these findings due to bias, confounding, or chance, although there may be residual confounding that we have not been able to fully address with the clinical and demographic variables that were measured. Furthermore, there may also be unmeasured confounding by lifestyle and other factors that are closely correlated with smoking, alcohol use, and obesity. Alternatively, these associations could have biological plausibility in this context and could represent a causal association.

Smoking

Our results for smoking are broadly consistent with 2 prior studies from Italy.16, 17 In a population-based study of 1138 patients diagnosed from 1991 to 1993 and followed through 2002, Battaglioli et al16 reported that the number of years smoked (HR, 1.26 for >39 years vs ≤25 years; 95% CI, 0.94-1.70) and the number of pack-years (HR, 1.60 for >31 pack-years vs ≤14 pack-years; 95% CI, 1.18-2.18) were associated with poorer survival after adjustment for gender, age, and education, although there was no association by smoking status (never, former, current) or years since quitting smoking. These associations were similar when the outcome was restricted to death due to lymphoma. In a second study of 268 NHL patients enrolled from 1983 through 2002, Talamini et al17 reported that the number of cigarettes per day (HR, 1.80 for ≥20 vs never smoked; 95% CI, 1.06-2.73) and, to a lesser extent, duration of smoking (HR, 1.32 for ≥30 years vs never smoked; 95% CI, 0.85-2.05) were associated with poorer survival after adjustment for age, sex, B-symptoms, and the IPI. With respect to NHL subtypes, the first study reported stronger associations for follicular lymphoma,16 whereas the second smaller study reported stronger associations for follicular lymphoma and CLL/SLL.17 These findings for follicular lymphoma are consistent with those of the current study; we had too few CLL/SLL cases in our study to evaluate survival.

Cigarette smoking could negatively influence NHL survival through multiple mechanisms,13, 25 including a direct carcinogen impact; activation of other carcinogens through induction of metabolizing enzymes; influence on immunologic function, including increasing proinflammatory cytokines and T-cell function; and impact on the prevalence of the t(14;18) translocation and subsequent overexpression of bcl-2, which can inhibit apoptosis. Smoking may also impact treatment efficacy, and smokers also tend to have more comorbidities that can also impact survival.25

Alcohol Use

Both of the Italian studies also reported positive associations with alcohol use. Compared with nondrinkers, drinking any alcohol (HR, 1.41; 95% CI, 1.10-1.81), drinking >32 versus ≤13 g of alcohol per day (HR, 1.41; 95% CI, 1.05-1.90), and longer duration of drinking (HR, 1.55 for >43 years vs ≤10 years; 95% CI, 1.13-2.14) were all associated with poorer survival in 1 study,16 whereas a greater number of drinks per day (HR, 1.69 for ≥4 vs <2 drinks; 95% CI, 1.04-2.76) was found to be associated with poorer survival in the other study.17 These associations were broadly elevated for the common subtypes. The poorer survival associated with any use of alcohol is consistent with the findings of the current study, although it must be noted that the Italian studies had a much higher level of alcohol use than the current one, raising concerns about the nature of any dose-response biologic effect. With respect to the risk of developing NHL, light to moderate alcohol use appears to be protective for NHL overall and for the major subtypes,14 which is the opposite of its association with outcome observed to date. The reasons for this are not clear, but if real, would suggest different biologic mechanisms for the 2 associations. Alcohol can act as both a direct and indirect carcinogen,26 and may also impact treatment efficacy.27 In contrast, alcohol has immunomodulatory effects28 that have been hypothesized to protect against the development of NHL at light to moderate levels of intake,29 complicating the interpretation of these findings.

Obesity

To the best of our knowledge, only a single study to date has evaluated the role of obesity in NHL prognosis. Tarella et al18 reported on 121 lymphoma patients enrolled in Italy from 1990 to 1997 who received high-dose chemotherapy and autograft and were followed for a median of 3 years. Patients with a BMI ≥28 kg/m2 had a poorer event-free (HR, 2.8; 95% CI, 1.5-5.3) and overall (RR, 2.9; 95% CI, 1.3-6.2) survival compared with patients with a BMI of <28 kg/m2 after adjustment for gender, IPI, bone marrow infiltration, and histology; histology-specific results were not provided.

In the current study, when smoking, alcohol use, and obesity were included in the same model that also adjusted for clinical and demographic factors, the associations were largely unchanged, suggesting that these may be independent prognostic factors. Both Battaglioli et al16 and Talamini et al17 reported a similar result when they simultaneously adjusted for smoking and alcohol use; however, to our knowledge, no other study to date has simultaneously assessed all 3 factors.

Obesity can affect cancer through multiple mechanisms, including an impact on estrogen, androgen, insulin, and growth factor pathways.30 Obesity is associated with a proinflammatory state in general and the release of tumor necrosis factor-α in particular, both of which have been suggested as NHL risk31, 32 and prognostic factors.10, 11 There is also the potential for underdosing of chemotherapy in obese individuals.33 Although all these factors can negatively impact existing comorbidities in NHL patients and therefore decrease overall survival, our finding that these associations remain when assessing lymphoma-specific survival suggest a more direct impact of these factors on lymphoma and/or its treatment.

Strengths and Limitations

Strengths of this study include the relatively large sample size, population-based case ascertainment, and relatively long-term and nearly complete follow-up. These results should be generalizable to community patients. There are also limitations. Data regarding smoking, alcohol use, and BMI were self-reported, but were reported before outcome. We also only collected smoking and alcohol data for a subset of cases, limiting the sample size for multivariate models. Alcohol use was light, and we could not assess heavier use such as that seen in the Italian studies.16, 17 We were unable to assess event-free survival or rarer NHL subtypes, or to adjust for standard prognostic factors such as the IPI or FLIPI, or detailed treatment. Nevertheless, these clinical and demographic factors have been shown to predict outcome similar to the IPI or FLIPI in this data set.10, 11 Additional limitations include the lack of data concerning changes in these factors after diagnosis; diagnosis and initial treatment of cases from 1998 through 2000, before the widespread use of rituximab as the standard of therapy for B-cell lymphoma34, 35; and a largely white study population. Future studies should address these limitations.

Conclusions

In conclusion, the current study findings of the adverse impact of smoking, alcohol use, and obesity are consistent with data from 3 smaller studies from Italy and, thus, support investigating whether smoking cessation, decrease in alcohol intake, or weight loss after a diagnosis of NHL can improve outcomes.

Acknowledgements

We thank Peter Hui for data management support, and Sondra Buehler for secretarial support.

CONFLICT OF INTEREST DISCLOSURES

Supported by the National Cancer Institute (Grants R01 CA96704 and P50 CA97274; NCI Intramural Program; and Surveillance, Epidemiology, and End Results contracts N01-PC-67,010, N01-PC-67,008, N01-PC-67,009, N01-PC-65,064, and N02-PC-71,105).

Ancillary