Cancer Epidemiology 39 (2015) 1093–1098

Contents lists available at ScienceDirect

Cancer Epidemiology
The International Journal of Cancer Epidemiology, Detection, and Prevention

journal homepage: www.cancerepidemiology.net

Risk of lymphoma subtypes and dietary habits in a Mediterranean area

Marcello Campagnaa , Pierluigi Coccoa,* , Mariagrazia Zuccab , Emanuele Angeluccic,
Attilio Gabbasa , Gian Carlo Latted , Antonella Urasd , Marco Raise , Sonia Sannab ,
Maria Grazia Ennasb
a
Department of Public Health, Clinical and Molecular Medicine, University of Cagliari, 09042 Monserrato, Cagliari, Italy
b
Department of Biomedical Sciences, Cytomorphology Section, University of Cagliari, 09042 Monserrato, Cagliari, Italy
c
Unit of Hematology, A. Businco Oncology Hospital, ASL 8, 09121 Cagliari, Italy
d
Unit of Hematology, San Francesco Hospital, ASL 3, 08100 Nuoro, Italy
e
Via Bonn 8, Cagliari 09129, Italy

A R T I C L E I N F O A B S T R A C T

Article history: Background: Previous studies have suggested that diet might affect risk of lymphoma subtypes. We
Received 9 November 2014 investigated risk of lymphoma and its major subtypes associated with diet in the Mediterranean island of
Received in revised form 8 August 2015 Sardinia, Italy.
Accepted 1 September 2015
Methods: In 1998–2004, 322 incident lymphoma cases and 446 randomly selected population controls
Available online 12 September 2015
participated in a case-control study on lymphoma etiology in central-southern Sardinia. Questionnaire
interviews included frequency of intake of 112 food items. Risk associated with individual dietary items
Keywords:
and groups thereof was explored by unconditional and polytomous logistic regression analysis, adjusting
Diet
Mediterranean diet
by age, gender and education.
Lymphoma, B-cell Results: We observed an upward trend in risk of lymphoma (all subtypes combined) and B-cell lymphoma
Lymphoma, non Hodgkin with frequency of intake of well done grilled/roasted chicken (p for trend = 0.01), and pizza (p for
Case-control study trend = 0.047), Neither adherence to Mediterranean diet nor a frequent intake of its individual
components conveyed protection. We detected heterogeneity in risk associated with several food items
and groups thereof by lymphoma subtypes although we could not rule out chance as responsible for the
observed direct or inverse associations.
Conclusions: Adherence to a Mediterranean diet does not seem to convey protection against the
development of lymphoma. The association with specific food items might vary by lymphoma subtype.
ã 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Diet might intervene in the development of lymphoma through

Abbreviations: BMI, body mass index; Cas, cases; CLL, chronic lymphocytic
leukaemia; ctrl, controls; DLBCL, diffuse large B-cell lymphoma; EPIC, European
prospective investigation of diet and cancer; FL, follicular lymphoma; HCAs,
heterocyclic amines; HL, Hodgkin lymphoma; MM, multiple myeloma; NHL, non
Hodgkin lymphoma; OR, odds ratio; 95% CI, 95% confidence intervals; PAHs,
polycyclic aromatic hydrocarbons; Q1–Q5, quantiles, approximate; WHO, World
Health Organization.
* Corresponding author at: Department of Public Health, Clinical and Molecular
Medicine, University of Cagliari, Asse didattico E, SS 554 km 4.500, 09042
Monserrato, Cagliari, Italy. Fax: +39 70 6754728.
E-mail addresses: makamp@libero.it (M. Campagna), coccop@medicina.unica.it
(P. Cocco), mzucca@unica.it (M. Zucca), emnang@tin.it (E. Angelucci),
atgabbas@tiscali.it (A. Gabbas), gclatte@gmail.com (G.C. Latte),
labematologia.hsf@aslnuoro.it (A. Uras), marcorais44@gmail.com (M. Rais),
soni.sa@hotmail.it (S. Sanna), makamp@libero.it (M.G. Ennas).
antigenic stimulation of lymphoid tissue in the gastrointestinal
tract due to he effect of specific nutrients, resulting in changes in
the immune system response [1]. However, in spite of a few reports
of an increasing risk [2,3], overall, results on red meat are not
convincing of its involvement in NHL risk, irrespective of cooking
method and grade of doneness, including those typically leading to
the formation of heterocyclic amines (HCAs) and polycyclic
aromatic hydrocarbons (PAHs) [4–8]. Likewise, intake of dairy
products has been reported as a risk factor in some studies [8,9],
but not in a large prospective trial conducted in Europe [7]. On the
other hand, subjects reporting a high fat fish intake run a lower risk
of NHL as well as all the other lymphatic cancer in a Canadian study
[10], while fresh fruits and vegetables might convey protection
[2,3,11,12].

http://dx.doi.org/10.1016/j.canep.2015.09.001
1877-7821/ ã 2015 Elsevier Ltd. All rights reserved.
1094 M. Campagna et al. / Cancer Epidemiology 39 (2015) 1093–1098
Because of its balanced supply of nutrients, the Mediterranean
diet is considered to prevent against chronic degenerative diseases,
including cardiovascular diseases and cancer [13]. However, in an
Italian case-control study, typical Mediterranean diet items, such
as pasta and rice, showed an association with NHL risk, while fruits
and vegetables showed an inverse association [14].
The Mediterranean island of Sardinia, Italy is home for the highest
rate of male centenarians in the world, and diet, together with genes
and physical activity, is thought to have contributed [15]. We used the
epidemiological data collected in a case-control study conducted in
central-southern Sardinia (Italy) in 1998–2004 to explore whether
food items typical of a Mediterranean diet might affect risk of
lymphoma and its most prevalent subtypes.
2. Materials and methods
Details on the study design can be found elsewhere [16]. Briefly,
incident cases were enrolled in two referral centres for diagnosis
and treatment of haematological malignancies in the region.
Histological diagnosis and slides were available for review for
70.1% of cases. A panel of pathologists (coordinated by MR)
reviewed the slides of 25% of the cases with available slides, to
apply the 2008 WHO classification of lymphomas. Overall, the
study population comprised 322 lymphoma cases (all subtypes
combined), including 186 B-cell lymphoma cases, 12 T-cell
lymphoma cases, 30 Hodgkin lymphoma cases, and 94 cases of
not otherwise specified NHL. For the latter cases, only the diagnosis
from the clinical records, and a first pathology report about the
morphological features, but not immunohistochemistry, was
available, because of dropout from the diagnostic work up. We
included these cases in the analysis of all lymphoma subtypes
combined. Controls were randomly selected within the same time
frame of case recruitment among the census population of the
study area, frequency matched to cases by gender, 5-year age-
groups, and area (local health unit) of residence. The study protocol
was approved by the Ethical Committee of the University Hospital
of Cagliari, and by the Ethical Committee of the Local Health Unit N.
3, Nuoro. All study participants signed an informed consent form
according to the Helsinki declaration. Participation rate was 88%
among the eligible cases, and to 69.2% among the eligible controls.
Trained interviewers used a semi-structured questionnaire elicit-
ing information on the frequency of intake of 112 dietary items,
categorized as never, occasional (up to once a month), 1–4 times a
month, 2–6 times a week, or daily. For eight type of fresh meat
items, preferred cooking method and degree of doneness,
classified as rare, medium, well done, and very well done (burned
on surface), were also asked. Beef, pork, horse meat and lamb were
analyzed separately, and combined within the commodity defini-
tion of red meat. We selected for analysis the most frequently
reported vegetable items, and we grouped separately leafy and
cruciferous vegetables. Onions and garlic were the most popular
condiments, and we analyzed them individually. A question on
olive oil was included in the dietary questionnaire when the study
was in the field, and the information was therefore available only
for a subset of the study population (42 cases and 112 controls). We
nevertheless mention this result, because of its a priori interest as a
typical component of the Mediterranean diet. We grouped
together citrus fruits, including oranges, tangerines, lemons and
grapefruits, and summer fruits, including cherries, nectarines,
plums, prunes, apricots, and strawberries. Nerve stimulant
beverages, such as coffee, tea, and cola beverages, were analyzed
individually.
Following Buckland et al. [17], we constructed a score of
adherence to a Mediterranean diet by summing up the frequency
score for each fresh fruit, vegetable, legume, fresh fish, shell fish
and crustacean (crabs, shrimps, crayfish, lobster and similar) in our
list of dietary items, along with pasta, rice, and bread. As a variation
in respect to the Buckland et al. score, we excluded white meat
(poultry and rabbit) from the calculation, and used the sum of
scores of the individual red meat items as a denominator, rather
than a subtrahend. Besides, we did not consider dairy food items
because of inconsistent opinions in the literature about these items
being components of the Mediterranean diet [17,18], and because
being historically an important part of the local diet and the local
economy. As it concerns alcohol, only a moderate daily intake of
wine, but not other alcoholic beverages, was considered to add to
the Mediterranean diet score, while abstinence from wine or a
more than moderate total alcohol intake did not add to the score.
Olive oil was also not included as pertinent information was
collected only in a fraction of the study population, and among
those who provided the information 81% were daily consumer,
with none having a less than weekly intake.
2.1. Statistical methods
We used unconditional logistic regression analysis to calculate
the odds ratio (OR) for lymphoma (all subtypes combined), B-cell
lymphoma overall, its most represented subtypes, including
DLBCL, CLL, FL, and MM, and HL, associated with frequency of
intake of selected food items and groups thereof, as well as the
cooking method and grade of doneness of meat items. To
categorize the intake of food groups, such as red meat, cruciferous
vegetables, leafy vegetables and citrus fruits, we summed up the
frequency score of each item in that group; then, we categorized
the sum into quartiles or quintiles based on its distribution among
the controls. For a few food groups, due to the concentration of
study subjects in a limited set of values of such sum of scores, when
a cut point between quantiles fell within one of such values,
moving to one quantile all the study subjects with the same score
generated a non homogeneous distribution of study subjects
across the quantile categories.
To account for changes in the age, gender and socio-economic
status balance between cases and controls, due to possibly
differential participation refusal rates, we adjusted the ORs by
age, as a continuous variables, gender and education (categorized
as primary school, middle school, high school or higher education).
Including other covariates, such as body mass index (BMI), alcohol
intake, or smoking in the regression models resulted in only
marginal changes in the risk estimates. We conducted polytomous
regression analysis and used the Cochran Q test to assess
heterogeneity in risk by gender and by histological subtype. We
also calculated two-tail 95% confidence intervals (95% CI) of the
ORs using Wald’s algorithm ðebðza=2 seb Þ Þ, and we used the Wald’s
test for trend to calculate the probability associated with a linear
increase or decrease in the ORs, as the ratio between the regression
coefficient associated with the continuous covariate of interest and
its standard error. We used the 5% (two-tail) threshold value to
reject chance.
The procedures followed in our study were in accordance with
the ethical standards of the responsible institutional and national
committee on human experimentation, and with the Helsinki
Declaration of 1975, as revised in 2000 and 2008.
3. Results
Table 1 summarizes the demographics of our study popula-
tions. Obese and overweight subjects and more than moderate
alcohol drinkers were more prevalent among the controls, and
current smokers were slightly more prevalent among the cases.
Most individual food items did not show an association with risk of
lymphoma (all subtypes combined). One exception was well done
 M. Campagna et al. / Cancer Epidemiology 39 (2015) 1093–1098 1095

Table 1 olive oil showed an inverse association of borderline significance

Selected characteristics of cases and controls participating to the study on the
(OR = 0.4, 95% 0.2–1.0; not shown in the tables). A high frequency
etiology of lymphoma in Sardinia, Italy.
intake of citrus fruits was also inversely associated with risk among
Cases Controls female study subjects (OR = 0.5, 95% 0.2–0.9), but not males
N (%) N (%)
(Supplementary Table 3).
Total 322 (100.0) 446 (100.0) Table 2 shows risk of lymphoma (all subtypes combined) by
Male 188 (58.0) 252 (56.5)
quantiles of intake of selected groups of food items. None showed
Female 134 (42.0) 194 (43.5)
significant associations nor trends in risk. In particular, we did not
Age (years) find an association with intake of well done or very well done
29 17 (5.3) 14 (3.1) grilled/roasted red meat, nor an inverse association with frequent
30–39 35 (10.9) 40 (9.0) intake of cruciferous or leafy vegetables or fruits. Adherence to a
40–49 44 (13.7) 63 (14.1)
typical Mediterranean diet also did not appear to convey
50–59 77 (23.9) 109 (24.4)
60–69 94 (29.2) 132 (29.6) protection. When limiting the analysis to B-cell lymphoma cases,
70 55 (17.1) 88 (19.7) risks associated with intake of red meat items were more elevated,
and trends of borderline significance for beef, lamb, and horse
Education (years)
meat, but no trend was observed for grilled/roasted red meat items
8 228 (70.8) 313 (70.2)
9 94 (29.2) 133 (29.8)
combined. For the rest of food items, there was only minimal
variation in the risk estimates (Supplementary Tables 4 and 5).
Body mass index When we conducted the analysis on NHL cases, thus excluding CLL,
Underweight (<18.5) 10 (3.0) 7 (1.5) MM, and HL from the case series, the associations we observed for
Regular (18.5–24.9) 166 (51.6) 176 (39.5)
lymphomas (all subtypes combined) and B-cell lymphomas
Overweight (25–29.9) 111 (34.5) 181 (40.6)
Obese (30) 35 (10.9) 82 (18.4) persisted substantially unchanged (not shown in the tables).
Table 3 shows risk of individual B-cell lymphoma subtypes and
Alcohol intake HL associated with high frequency intake of selected food items
Abstinent 139 (43.3) 184 (41.4) with reference to the lowest category. A high frequency intake of
Moderate (60 g per day or less) 95 (29.4) 132 (29.5)
High (more than 60 g per day) 88 (27.2) 130 (29.1)
red meat items showed an association with risk of DLBCL and CLL,
but not FL, MM, and HL. Cooking method and grade of doneness did
Smoking not make a difference. An increase in risk of all subtypes, but MM,
Non smokers 126 (39.1) 185 (41.5) was observed with a frequent intake of well done and very well
Ex smokers 74 (23.0) 94 (21.1)
done grilled/roasted chicken. A daily intake of pasta was a risk
Current smokers 122 (37.9) 165 (37.0)
Missing 0 4 (0.4) factor for CLL and MM, while having a pizza more than once a week
was related to an apparent increase in risk of CLL. As it concerns the
rest of food items, a high frequency intake of tomatoes was
inversely associated with risk of DLBCL, and associated with an
and very well done grilled/roasted chicken, which daily intake was increased risk of CLL; a high intake of cruciferous vegetables was
associated with a 80% excess risk (OR = 1.8, 95% CI 1.2–2.9), with a inversely associated with MM risk only; olive oil appeared
significant upward trend by frequency of intake (p = 0.01) Risks inversely related to risk of any lymphoma subtype. Risk associated
were more elevated and the trend confirmed among women, but with high frequency intake of the rest of food items showed only
not men (Q test for heterogeneity = 3.161; p = 0.075; Supplemen- random variations across lymphoma subtypes. Adherence to a
tary Table 3). A second exception was pizza, which frequent Mediterranean diet was inversely associated with risk of DLBCL
(2–6 times per week) intake was also associated with an elevated (highest vs lowest quartile, OR = 0.4, 95% CI 0.1, 1.0; Wald’s test for
risk (OR = 1.8, 95% 1.1–3.1) and a borderline significant upward trend –2.33; p = 0.020), with not enough statistical power to test
trend (p = 0.047) (Supplementary Tables 1 and 2). A daily intake of the observed inverse association for HL. We observed significant

Table 2
Risk of lymphoma (all subtypes combined) associated with quantiles of intake of grouped food items.

Quantile of intake

Food items Q1 Q2 Q3 Q4 Q5

Cas/ctrl OR (95% CI) Cas/ctrl OR (95% CI) Cas/ctrl OR (95% CI) Cas/ctrl OR (95% CI) Cas/ctrl OR (95% CI)
Well done and very well done, 68/99 1.0 62/93 0.9 (0.6–1.5) 71/83 1.1 (0.7–1.8) 47/90 0.7 (0.4–1.1) 74/81 1.3 (0.8–2.0)
grilled/ roasted red meata
Processed meatc 78/103 1.0 66/127 0.7 (0.4–1.0) 24/36 0.8 (0.4–1.5) 70/91 1.0 (0.6–1.5) 84/89 1.1 (0.7–1.8)
Fresh seafoodb 41/58 1.0 82/112 1.0 (0.6–1.6) 64/87 1.0 (0.6–1.6) 87/106 1.1 (0.6–1.7) 49/84 0.8 (0.4–1.3)
Cheese 79/130 1.0 122/158 1.3 (0.9–1.9) 36/59 1.0 (0.6–1.6) 82/88 1.5 (1.0–2.3)
Cruciferous vegetables 128/177 1.0 69/86 1.1 (0.8–1.7) 18/27 1.0 (0.5–1.9) 107/156 1.0 (0.7–1.4)
Leafy vegetables 103/121 1.0 66/107 0.7 (0.5–1.1) 31/46 0.8 (0.5–1.4) 122/172 0.9 (0.6–1.2)
Summer fruits 93/119 1.0 99/133 0.9 (0.6–1.4) 122/178 0.8 (0.6–1.2) 8/16 0.6 (0.3–1.6)
Citrus fruits 106/160 1.0 46/83 0.8 (0.5–1.3) 124/116 1.6 (1.1–2.3) 46/87 0.8 (0.5–1.3)
Mediterranean diet score 63/92 1.0 80/71 1.7 (1.1–2.7) 60/94 1.0 (0.6–1.5) 63/91 1.1 (0.7–1.7) 57/99 0.9 (0.6–1.5)

Categories are based on quintiles for red meat, processed meat, seafood, and the Mediterranean diet score, and quartiles for cheese, cruciferous vegetables, leafy vegetables,
summer fruits and citrus fruits. For several items, the score distribution was far from being uniform across quantiles; using the cut points among quantiles generated
differences in the cell size.
Abbreviations: Q1–Q5, increasing quantiles of intake; cas, cases; ctrl, control; 95% CI, 95% confidence interval.
a
The commodity definition of red meat includes beef, pork, lamb and horse meat.
b
Fresh seafood includes fresh fish, shellfish and crustaceans.
c
Processed meat includes all kinds of cured meat (prosciutto, salami, sausage, bologna, and many others) typically served raw.
 1096
Table 3
Risk (OR and 95% CI) of B-cell lymphoma, its most represented mature subtypes (DLBCL, FL, CLL and MM), and Hodgkin Lymphoma (HL) associated with the highest category of frequency or quantile of intake of selected food items
with reference to the lowest category or quantile. The Q test for heterogeneity across the specific lymphoma subtypes is presented in the rightmost column.

Lymphoma subtype

Food items B-cell lymphoma DLBCL CLL FL Multiple myeloma Hodgkin lymphoma Q test
(p-value)
Cas/ctrl OR 95% CI Cas/ctrl OR 95% CI Cas/ctrl OR 95% CI Cas/ctrl OR 95% CI Cas/ctrl OR 95% CI Cas/ctrl OR 95% CI
Meat, any cooking method and grade of doneness
Beef 141/302 1.5 (1.0–2.2) 46/302 1.9 (0.9–3.7) 48/302 1.8 (1.0–3.4) 15/302 0.8 (0.3–1.9) 16/302 1.0 (0.4–2.6) 23/302 1.4 (0.6–3.6) 3.317 (0.506)
Lamb 93/172 1.5 (1.1–2.2) 32/172 2.0 (1.1–3.6) 32/172 1.7 (1.0–3.0) 11/172 1.3 (0.6–3.1) 8/172 0.6 (0.2–1.6) 14/172 1.4 (0.6–3.2) 4.855 (0.302)
Horse meat 93/193 1.4 (0.9–2.1) 35/193 1.9 (1.1–3.4) 31/193 1.5 (0.8–2.5) 12/193 1.3 (0.6–3.1) 7/193 0.4 (0.2–1.1) 16/193 1.1 (0.5–2.4) 7.035 (0.134)

M. Campagna et al. / Cancer Epidemiology 39 (2015) 1093–1098

Chicken 121/280 1.1 (0.8–1.6) 37/280 1.0 (0.5–1.7) 42/280 1.4 (0.8–2.5) 16/280 1.2 (0.5–2.9) 15/280 0.9 (0.4–2.2) 23/280 1.4 (0.5–3.4) 1.219 (0.875)

Well done and very well done grilled/roasted meat

Red meat 70/146 1.4 (0.8–2.4) 24/146 2.4 (0.9–6.6) 26/146 1.8 (0.8–4.2) 9/146 1.5 (0.4–5.9) 3/146 0.2 (0.1–0.9) 12/146 1.5 (0.4–5.0) 8.158 (0.086)
Chickenb 34/60 2.0 (1.2–3.4) 12/60 2.9 (1.2–6.9) 12/60 1.9 (0.9–4.2) 4/60 1.5 (0.4– 5.1) 2/60 0.9 (0.2–4.8) 7/60 2.6 (0.8–8.4) 1.925 (0.750)

Other food items

Pasta 53/103 1.3 (0.9–1.9) 9/103 0.6 (0.3–1.2) 25/103 2.4 (1.3–4.2) 3/103 0.5 (0.1–1.6) 11/103 2.8 (1.2–6.7) 9/103 1.1 (0.4–2.5) 14.69 (0.005)
Pizza 71/160 1.9 (1.0–3.5) 29/160 1.3 (0.5–3.5) 19/160 2.5 (0.8–7.3) 8/160a 9/160 0.7 (0.2–2.8) 20/160 0.9 (0.2–3.5) 2.380 (0.497)
Fresh seafood 33/84 0.9 (0.5–1.6) 15/84 0.9 (0.4–2.2) 11/84 1.5 (0.5–4.4) 3/84 2.1 (0.2–21.1) 2/84 0.3 (0.1–1.4) 5/84 1.3 (0.2–7.5) 3.658 (0.454)
Milk 113/261 1.2 (0.8–1.8) 32/261 1.1 (0.5–2.1) 36/261 1.1 (0.6–2.0) 16/261 1.6 (0.6–4.5) 16/261 1.2 (0.5–3.0) 21/261 2.0 (0.7–5.8) 1.554 (0.817)
Cheese 55/99 1.5 (0.9–2.4) 19/99 1.2 (0.6–2.5) 16/99 1.1 (0.5–2.3) 10/99 2.2 (0.8–6.3) 5/99 0.9 (0.2–3.3) 10/99 2.2 (0.7–6.8) 2.325 (0.676)
Cruciferous vegetables 68/157 1.2 (0.7–2.1) 25/157 1.5 (0.6–3.4) 19/157 1.4 (0.5–3.7) 14/157 6.3 (0.8–48.7) 2/157 0.1 (0.01–0.6) 10/157 2.2 (0.6–9.1) 11.03 (0.026)
Leafy vegetables 78/173 1.0 (0.6–1.6) 27/173 0.7 (0.3–1.4) 26/173 2.2 (0.8–2.1) 11/173 1.1 (0.3–3.6) 7/173 0.9 (0.2–3.6) 8/173 0.6 (0.2–1.8) 4.440 (0.350)
Tomatoes 22/59 1.0 (0.5–1.9) 44/386 0.5 (0.2–1.0) 59/386 3.3 (1.0–11.1) 19/386 0.6 (0.2–1.7) 23/386 3.4 (0.4–25.8) 23/386 0.5 (0.2–1.2) 10.66 (0.031)
Olive oil 18/95 0.4 (0.1–1.3) 5/95a 3/95 0.2 (0.03–1.7) 1/95 0.1 (0.01–1.1) 6/95 0.7 (0.1–6.9) 1/95 0.1 (0.01–0.6) 2.640 (0.451)
Citrus fruit 32/87 1.2 (0.7–2.0) 33/203 1.6 (0.9–2.9) 41/203 2.4 (1.4–4.3) 16/203 2.4 (1.0–5.8) 7/203 0.4 (0.2–1.1) 14/203 1.2 (0.5–2.6) 11.17 (0.025)
Apples 17/48 1.0 (0.5–1.9) 5/48 0.8 (0.3–2.3) 6/48 1.4 (0.4–4.5) 1/48 0.6 (0.1–5.5) 3/48 1.2 (0.3–5.8) 0/48a 0.873 (0.832)
Pears 138/324 0.8 (0.5–1.3) 37/324 0.7 (0.4–1.3) 50/324 1.5 (0.8–2.9) 19/324 1.4 (0.5–3.9) 19/324 1.3 (0.5–3.6) 20/324 0.9 (0.4–2.1) 3.243 (0.518)
Coffee 137/310 1.3 (0.9–1.9) 39/310 1.0 (0.5–1.8) 46/310 1.4 (0.8–2.6) 19/310 1.7 (0.6–4.8) 18/310 1.6 (0.6–4.4) 22/310 1.3 (0.5–3.1) 1.403 (0.844)

Dietary pattern
Mediterranean diet score 32/99 0.8 (0.5–1.4) 6/99 0.4 (0.1–1.0b) 13/99 1.0 (0.4–2.4) 4/99 1.9 (0.3–10.8) 6/99 0.9 (0.3–3.1) 2/99 0.3 (0.1–1.8) 4.428 (0.351)

The reference category was “never” for pizza and mozzarella, and well done roasted chicken; less than weekly for beef, pork, horse meat, chicken, tomatoes, apples, and pears; less than daily for pasta, olive oil, and coffee; the lowest
quantile for fresh seafood, cheese, cruciferous vegetables, leafy vegetables, and the adherence to Mediterranean diet score; the two lowest quintiles combined for well done and very well done red meat.
a
Fisher’s exact test: DLBCL and daily intake of olive oil: p = 0.317; FL and pizza p = 0.739; HL and apples: p = 0.902.
b
Wald’s test for trend: DLBCL –2.326, p = 0.020; HL = 0.999, p = 0.318.
 M. Campagna et al. / Cancer Epidemiology 39 (2015) 1093–1098
heterogeneity in risk by subtypes for pasta (p = 0.005), cruciferous
vegetables (p = 0.026), tomatoes (p = 0.031), and citrus fruit
(p = 0.025).
4. Discussion
Consistently with a previous report on the heterogeneity of
lymphoma subtypes in respect to the epidemiological associations
[19], we detected a significant heterogeneity in risk associated
with a daily intake of pasta, tomatoes and citrus fruit by lymphoma
subtype, with CLL more frequently showing significant excess
risks. Adherence with a Mediterranean diet did not convey
protection against lymphoma overall and most of its subtypes,
with the remarkable exception of DLBCL, which risk showed a
significantly inverse trend. Our results also tentatively suggested
an elevated risk of lymphoma (all subtypes combined) and B-cell
lymphoma with high frequency intake of well done and very well
done grilled/roasted chicken, and an inverse association with citrus
fruits, among females, but not in male study subjects. Results of the
European prospective investigation of diet and cancer (EPIC) also
showed an association between a high intake of poultry and risk of
B-cell lymphoma [7]. Risk did not vary by intake of dairy and
seafood.
Swedish and US studies reported an elevated NHL risk
associated with a high frequency intake of fried red meat animal
protein and dairy food items [2,3], and an inverse association with
fruits and vegetables [2,3,11], particularly with leafy and crucifer-
ous vegetables [11]. The associations were mainly observed among
women [2,3]. In our study, frequency of intake of leafy and
cruciferous vegetables did not affect risk of lymphoma (all
subtypes combined) and B-cell lymphoma. Among subtypes, we
noted an inverse association of frequent intake of cruciferous
vegetables with MM risk only.
We also observed an increase in risk of lymphoma overall and B-
cell lymphoma, but not of histologically confirmed subtypes, with
increasing frequency of intake of pizza, and an apparent protective
effect of a daily intake of olive oil, equally shared by gender and
lymphoma subtypes. As it concerns pizza, this is notoriously a
combination of fresh dough typically topped with tomato sauce,
mozzarella and basil, and variable items in a customized fashion,
baked at very high temperature, traditionally in a wood fired oven.
The surface is frequently burned, and it is therefore considered one
of the major dietary sources of polycyclic aromatic hydrocarbons
(PAH) [20], although it is unclear whether ingested or inhaled PAH
would contribute to lymphomagenesis.
In a Canadian study, the fraction of daily caloric intake from
seafood showed a dose-related inverse trend with risk of NHL, MM,
and leukaemia [10]. More than two thirds of our study population
reported to have fresh fish meals twice a week or more (527/768,
68.9%); we did not observe an association with lymphoma (all
subtypes combined) nor B-cell lymphoma; among subtypes, risk of
MM was the only showing an inverse association with high
frequency intake of seafood. Similar findings have been previously
published [21–24].
In two studies conducted in Northern Italy, NHL risk increased
with high frequency intake of pasta, rice and cheese, and it decreased
with high frequency intake of eggs, fruits and vegetables [14].
Associations between dietary items and nutrients and risk of specific
lymphoma subtypes have been reported in other studies with
inconsistent findings [25,26], while studies of HL have rarely
addressed dietary habits, with no indication of an association [21,27].
We are aware that of several issues that limit interpretation of
our findings:
1. It is quite possible that both the observed increases and
decreases in risk were due to the random variation in the risk
1097
estimates around the null, particularly in the analyses by gender
and by subtype, as a consequence of the small study size and the
elevated number of comparisons we made.
2. We did not adjust our risk estimates by BMI, alcohol or smoking.
However, introducing these covariates in the regression models
resulted in only marginal changes in the risk estimates (not
shown in the tables) and previous large studies and pooled
analyses did not confirm an association with obesity [28],
smoking, or alcohol [29].
3. Our dietary section in the questionnaire was a compromise
aiming to extend the field of investigation while avoiding an
excessively long interview, due to the multiple etiological
hypotheses being addressed. Therefore, we did not use a food
atlas or put specific questions that would have allowed to
estimate the portion size, and the intake of specific nutrients.
4. The accuracy of self reports of dietary habits 2–3 years before
interview might be questionable. We did not pursue any
attempt to validate the questionnaire information. However, as
diet is not commonly considered to be a risk factor for
lymphoma, we are confident that inaccuracy in providing
information was not differentially distributed by case-control
status. Such non differential inaccuracy might have lead to
underestimating direct and inverse associations, if any. For all
the above reasons, our analyses are to be considered as
exploratory, not allowing any conclusive remark.
On the other hand, previous studies did not explore adherence
to the Mediterranean diet, which is considered to prevent
cardiovascular diseases and cancer [13], nor did they assess
heterogeneity in risk associated with dietary items by gender or
lymphoma subtype. Still, we have to mention that our way of
calculating a score of adherence to the Mediterranean diet differed
from the one proposed by Buckland et al. [17] and effectively tested
in the EPIC cohort study on coronary artery disease, and it was not
validated. Although we motivated the changes we made, this adds
to concern in interpreting our findings. However, when we run the
analysis of risk associated with adherence to the Mediterranean
diet using the algorithm proposed by Buckland et al., the overall
pattern of no association with risk was confirmed (not shown in
the tables).
5. Conclusions
In conclusion, our results suggest that adherence to a
Mediterranean diet and frequency of intake of its individual
components do not seem to convey protection against the
development of lymphoma (all subtypes combined) and B cell
lymphoma. We cannot exclude chance as responsible for the
heterogeneity we observed in the associations by lymphoma
subtype.
Author contributions
PC designed the study. MZ, SS, AU and MGE conducted the
laboratory analyses. MC managed the data and conducted the
epidemiological analysis. MC and PC wrote, and MZ, EA, AG, GCL,
MR, and MGE reviewed the manuscript. All authors made
significant contributions to, reviewed, and approved the final
version of the manuscript
Conflict of interest
None of the co-authors declare financial relationships with any
organization that might have an interest in the submitted work,
nor other relationships or activities that could appear to have
influenced the submitted work. The institutions providing
1098 M. Campagna et al. / Cancer Epidemiology 39 (2015) 1093–1098
financial support did not intervene or influence in any way the
recruitment of study subjects, or the writing of the manuscript, or
the decision of submitting it for publication
Acknowledgments
Funding for this study was provided by Italian Association for
Cancer Research (AIRC, Investigator Grant 11855), PI: Pierluigi
Cocco; Fondazione Banco di Sardegna 2010–2012, PI Maria Grazia
Ennas; and Regione Autonoma della Sardegna (LR7CRP-59812/
2012), PI Maria Grazia Ennas.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.
canep.2015.09.001.
References
[1] L. Hoyles, J. Vulevic, Diet, immunity and functional foods, Adv. Exp. Med. Biol.
635 (2008) 79–92.
[2] E.T. Chang, K.E. Smedby, S.M. Zhang, H. Hjalgrim, M. Melbye, A. Ost, et al.,
Dietary factors and risk of non-Hodgkin lymphoma in men and women, Cancer
Epidemiol. Biomark. Prev. 14 (2005) 512–520.
[3] T. Zheng, T.R. Holford, B. Leaderer, Y. Zhang, S.H. Zahm, S. Flynn, et al., Diet and
nutrient intakes and risk of non-Hodgkin’s lymphoma in Connecticut women,
Am. J. Epidemiol. 159 (2004) 454–466.
[4] M.H. Ward, S.H. Zahm, D.D. Weisenburger, G. Gridley, K.P. Cantor, R.C. Saal,
et al., Dietary factors and non-Hodgkin’s lymphoma in Nebraska (United
States), Cancer Causes Control 5 (1994) 422–432.
[5] A.J. Cross, M.H. Ward, M.J. Schenk, M. Kulldorff, W. Cozen, S. Davis, et al., Meat
and meat-mutagen intake and risk of non-Hodgkin lymphoma: results from a
NCI-SEER case-control study, Carcinogenesis 27 (2006) 293–297.
[6] C.R. Daniel, R. Sinha, Y. Park, B.I. Graubard, A.R. Hollenbeck, L.M. Morton, et al.,
Meat intake is not associated with risk of non-Hodgkin lymphoma in a large
prospective cohort of U.S. men and women, J. Nutr. 142 (2012) 1074–1080.
[7] S. Rohrmann, J. Linseisen, M.U. Jakobsen, K. Overvad, O. Raaschou-Nielsen, A.P.
Tjonneland, et al., Consumption of meat and dairy and lymphoma risk in the
European Prospective Investigation into Cancer and Nutrition, Int. J. Cancer
128 (2011) 623–634.
[8] B. Charbonneau, H.M. O’Connor, A.H. Wang, M. Liebow, C.A. Thompson, Z.S.
Fredericksen, et al., Trans fatty acid intake is associated with increased risk and
n3 fatty acid intake with reduced risk of non-Hodgkin lymphoma, J. Nutr. 143
(2013) 672–681.
[9] N.J. Ollberding, B. Aschebrook-Kilfoy, D.B. Caces, M.E. Wright, D.D.
Weisenburger, S.M. Smith, et al., Phytanic acid and the risk of non-Hodgkin
lymphoma, Carcinogenesis 34 (2013) 170–175.
[10] L. Fritschi, G.L. Ambrosini, E.V. Kliewer, K.C. Johnson, Dietary fish intake and
risk of leukaemia, multiple myeloma, and non-Hodgkin lymphoma, Cancer
Epidemiol. Biomark. Prev. 13 (2004) 532–537.
[11] L.E. Kelemen, J.R. Cerhan, U. Lim, S. Davis, W. Cozen, M. Schenk, et al.,
Vegetables, fruit, and antioxidant-related nutrients and risk of non-Hodgkin
lymphoma: a National Cancer Institute—Surveillance, Epidemiology, and End
Results population-based case-control study, Am. J. Clin. Nutr. 83 (2006) 1401–
1410.
[12] B.C.H. Chiu, S. Kwon, A.M. Evens, T. Surawicz, S.M. Smith, D.D. Weisenburger,
Dietary intake of fruit and vegetables and risk of non-Hodgkin lymphoma,
Cancer Causes Control 22 (2011) 1183–1195.
[13] E. Gotsis, P. Anagnostis, A. Mariolis, A. Vlachou, N. Katsiki, A. Karagiannis,
Health benefits of the Mediterranean diet: an update of research over the last
5 years, Angiology (2014) (Epub ahead of print).
[14] R. Talamini, J. Polesel, M. Montella, L. Dal Maso, M. Crovatto, A. Crispo, et al.,
Food groups and risk of non-Hodgkin lymphoma: a multicenter, case-control
study in Italy, Int. J. Cancer 118 (2006) 2871–2876.
[15] R. Koenig, Sardinia’s mysterious male Methuselah, Science 291 (2001)
2074–2076.
[16] P. Cocco, G. Piras, M. Monne, A. Uras, A. Gabbas, M.G. Ennas, et al., Risk of
malignant lymphoma following viral hepatitis infection, Int. J. Hematol. 87
(2008) 474–483.
[17] G. Buckland, C.A. Gonzalez, A. Agudo, M. Vilardell, A. Berenguer, P. Amiano,
et al., Adherence to the Mediterranean diet and risk of coronary heart disease
in the Spanish EPIC cohort study, Am. J. Epidemiol. 170 (2009) 1518–1529.
[18] W.C. Willett, F. Sacks, A. Trichopoulou, G. Drescher, A. Ferro-Luzzi, E. Helsing,
et al., Mediterranean diet pyramid: a cultural model for healthy eating, Am. J.
Clin. Nutr. 61 (Suppl) (1995) 1402S–1406S.
[19] L.M. Morton, S.L. Slager, J.R. Cerhan, S.S. Wang, C.M. Vajdic, C.F. Skibola, et al.,
Etiologic heterogeneity among non-Hodgkin lymphoma subtypes: the
InterLymph Non-Hodgkin Lymphoma Subtypes Project, J. Natl. Cancer Inst.
Monogr. 48 (2014) 130–144.
[20] M. Lodovici, P. Dolara, C. Casalini, S. Ciappellano, G. Testolin, Polycyclic
aromatic hydrocarbon contamination in the Italian diet, Food Addit. Contam.
12 (1995) 703–713.
[21] E. Fernandez, L. Chatenoud, C. La Vecchia, E. Negri, S. Franceschi, Fish
consumption and cancer risk, Am. J. Clin. Nutr. 70 (1999) 85–90.
[22] L.M. Brown, G. Gridley, L.M. Pottern, D. Baris, C.A. Swanson, D.T. Silverman,
et al., Diet and nutrition as risk factors for multiple myeloma among blacks and
whites in the United States, Cancer Causes Control 12 (2001) 117–125.
[23] L. Fritschi, G.L. Ambrosini, E.V. Kliewer, K.C. Johnson, Dietary fish intake and
risk of leukaemia, multiple myeloma, and non-Hodgkin lymphoma, Cancer
Epidemiol. Biomark. Prev. 13 (2004) 532–537.
[24] H.D. Hosgood 3rd, D. Baris, S.H. Zahm, T. Zheng, A.J. Cross, Diet and risk of
multiple myeloma in Connecticut women, Cancer Causes Control 18 (2007)
1065–1076.
[25] C.A. Thompson, T.M. Habermann, A.H. Wang, R.A. Vierkant, A.R. Folsom, J.A.
Ross, et al., Antioxidant intake from fruits, vegetables and other sources and
risk of non-Hodgkin’s lymphoma: the Iowa Women’s Health Study, Int. J.
Cancer 126 (2010) 992–1003.
[26] B. Aschebrook-Kilfoy, N.J. Ollberding, C. Kolar, T.A. Lawson, S.M. Smith, D.D.
Weisenburger, et al., Meat intake and risk of non-Hodgkin lymphoma, Cancer
Causes Control 23 (2012) 1681–1692.
[27] A. Tavani, C. La Vecchia, S. Gallus, P. Lagiou, D. Trichopoulos, F. Levi, et al.,
Red meat intake and cancer risk: a study in Italy, Int. J. Cancer 86 (2000)
425–428.
[28] E.V. Willett, L.M. Morton, P. Hartge, N. Becker, L. Bernstein, P. Boffetta, et al.,
Non-Hodgkin lymphoma and obesity: a pooled analysis from the InterLymph
consortium, Int. J. Cancer 122 (2008) 2062–2070.
[29] H. Besson, P. Brennan, N. Becker, A. Nieters, S. De Sanjosé, R. Font, et al.,
Tobacco smoking, alcohol drinking and non-Hodgkin’s lymphoma: an
European multi-centre case-control study (Epilymph), Int. J. Cancer 119
(2006) 901–908.