Cancer Causes & Control

, Volume 19, Issue 10, pp 1209–1215

Dietary intake of carotenoids and retinol and endometrial cancer risk in an Italian case–control study

Authors

    • Istituto di Ricerche Farmacologiche “Mario Negri”
  • Luigino Dal Maso
    • Unità di Epidemiologia e BiostatisticaCentro di Riferimento Oncologico
  • Maurizio Montella
    • Unità di EpidemiologiaIstituto Nazionale Tumori “Fondazione Giovanni Pascale”
  • Maria Parpinel
    • Istituto di Igiene ed Epidemiologia, DPMSCUniversità degli Studi di Udine
  • Eva Negri
    • Istituto di Ricerche Farmacologiche “Mario Negri”
  • Renato Talamini
    • Unità di Epidemiologia e BiostatisticaCentro di Riferimento Oncologico
  • Aldo Giudice
    • Unità di EpidemiologiaIstituto Nazionale Tumori “Fondazione Giovanni Pascale”
  • Silvia Franceschi
    • International Agency for Research on Cancer
  • Carlo La Vecchia
    • Istituto di Ricerche Farmacologiche “Mario Negri”
    • Istituto di Statistica Medica e Biometria “G. A. Maccacaro”Università degli Studi di Milano
Original Paper

DOI: 10.1007/s10552-008-9190-1

Cite this article as:
Pelucchi, C., Dal Maso, L., Montella, M. et al. Cancer Causes Control (2008) 19: 1209. doi:10.1007/s10552-008-9190-1

Abstract

Objectives

To provide information on the relation between intake of carotenoids and retinol and endometrial cancer, since available data are inconsistent. Further, carotenoids other than beta-carotene have been rarely investigated.

Methods

We conducted a multi-centric case–control study in various areas of Italy between 1992 and 2006 on 454 women with incident, histologically confirmed endometrial cancer and 908 controls admitted to the same network of hospitals of cases for acute, non-neoplastic conditions. Intake of carotenoids and retinol was computed from a validated and reproducible food frequency questionnaire. We adjusted for selected covariates, including energy intake, and calculated multivariate odds ratios (OR) and 95% confidence intervals (CI) using conditional logistic regression.

Results

Comparing the highest to the lowest quartile of intake, the ORs of endometrial cancer were 0.69 (95% CI, 0.48–0.99) for beta-carotene, 0.65 (95% CI, 0.45–0.94) for beta-cryptoxanthin, and 0.59 (95% CI, 0.41–0.85) for lutein plus zeaxanthin intake. No association emerged with retinol (OR = 1.31, 95% CI, 0.94–1.84), alpha-carotene (OR = 0.94, 95% CI, 0.66–1.34), and lycopene (OR = 0.95, 95% CI, 0.68–1.34).

Conclusions

Our results support a favorable role of selected dietary carotenoids on endometrial cancer risk.

Keywords

CarotenoidsCase–control studyDietEndometrial neoplasmsEpidemiologic studies

Introduction

A few epidemiological studies have considered the issue of retinoids and carotenoids and endometrial cancer [110], with interest focused chiefly on retinol and beta-carotene. Though the majority of studies found a moderate inverse relation between retinol intake and endometrial cancer [14], data are not consistent [5]. Similarly, risks were significantly reduced for a high intake of beta-carotene in three investigations from Western New York (OR = 0.4), Italy (OR = 0.5), and Shanghai (OR = 0.6) [35], but not in studies from Canada (OR = 1.0) and Mexico (OR = 0.8) [1, 7].

The intake of other carotenoids was less frequently examined. McCann et al. [3] reported that high intake of alpha-carotene and of lutein plus zeaxanthin decreased the risk of endometrial cancer by 40 and 70%, respectively. Lutein alone was however unrelated with risk in another study (OR = 0.80 for the highest versus lowest quartile of intake) [7]. Cryptoxanthin intake was considered in two studies, and both found no association with endometrial cancer [3, 7]. On the other hand, promising findings have been reported for lycopene, i.e., a non-provitamin A carotenoid. A high lycopene intake decreased the risk of endometrial cancer by 15% in a Canadian study, and by 40% in the Western New York Diet Study [3, 7]. These results were confirmed in a case–control analysis nested in a Canadian cohort, that reported a significant inverse relation with lycopene (hazard ratio, HR = 0.63, 95% CI, 0.43–0.94) [6]. In the latter study, no important associations emerged with any of the other carotenoids considered, including beta-carotene, lutein, and cryptoxanthin, nor with vitamin A.

Several plausible explanations for an anticancer role of carotenoids have been proposed, including their antioxidant effects, the regulation of cell differentiation via their conversion to vitamin A, and a modulation of immunologic functions [11].

Therefore, we examined the potential relation between retinol, selected carotenoids and the risk of endometrial cancer in a multi-centric case–control study conducted in Italy.

Materials and methods

The data were derived from a case–control study of endometrial cancer conducted between 1992 and 2006 in three Italian areas: the provinces of Pordenone and Milan, in northern Italy, and of Naples, in southern Italy [12].

Cases were 454 women (median age 60 years, range:18–79) with incident, histologically confirmed endometrial cancer, and no previous diagnosis of cancer. Controls were 908 women (median age 61 years, range:19–79) admitted to the same network of hospitals of cases for a wide spectrum of non-neoplastic acute illnesses. Women admitted for gynecological or hormone-related conditions, or any medical condition associated with long-term dietary changes were not eligible as controls. Women with a history of hysterectomy were excluded from the control group. Controls were admitted for traumas (36%), other orthopedic disorders (32%), acute surgical conditions (9%), and miscellaneous other illnesses, including eye, nose, ear, skin, or dental disorders (23%). Controls were frequency matched to cases with a 2:1 ratio on age and study center. Less than 5% of cases and controls approached refused to be interviewed.

Trained interviewers visited the hospital at regular intervals, and collected data during the hospital stay, using a structured questionnaire. This included information on socio-demographic characteristics, anthropometric measures, occupational and recreational physical activity, selected lifestyle habits, including tobacco smoking and alcohol drinking, a personal medical history, family history of cancer, menstrual and reproductive factors, history of oral contraceptive (OC) and hormone replacement therapy (HRT) use, and diet. A food frequency questionnaire (FFQ) was used to assess the usual diet during the two years preceding cancer diagnosis or hospital admission (for controls), to estimate the intake of total energy and of selected nutrients. The questionnaire included 78 foods, food groups, or dishes divided into six sections: (1) bread, cereals, first courses; (2) second courses (i.e., meat, fish, and other main dishes); (3) side dishes (i.e., vegetables); (4) fruits; (5) sweets, desserts, and soft drinks; (6) milk, hot beverages, and sweeteners. Study subjects were asked to indicate the average weekly frequency of consumption and, for about half of the items (including all questions on vegetables), their usual portion size. Intakes lower than once a week, but at least once a month, were coded as 0.5 per week. For a few vegetables and fruits, seasonal consumption and the corresponding duration were elicited. At the end of each section, one or two open questions were used to include other foods eaten at least once per week. Several questions aimed at assessing fat intake pattern were also included in the questionnaire. Dietary supplements were not considered, given their low frequency of consumption in this population [13, 14]. The intake of energy and of selected macro- and micronutrients was estimated using an Italian food composition database [15]. Both data on frequency of consumption and portion size were used in the calculation procedure. The intake of beta-carotene equivalents was defined as the sum of the intake of beta-carotene and half intake of alpha-carotene, alpha- and beta-cryptoxanthin. However, its value does not correspond to the exact sum, as it derived from different sources [16]. Thirty-five out of 78 items of the FFQ contributed to the calculation of intake of retinol and 63 of various carotenoids. A number of other foods, collected through the open questions, contributed to a lower extent to total retinol and carotenoids intake. The FFQ was reproducible and satisfactorily valid [17, 18]. The Pearson correlation coefficients were 0.70 for reproducibility and 0.61 for validity of information on energy intake. The correlation coefficients for beta-carotene equivalents were 0.62 for reproducibility and 0.43 (adjusted for total energy, age, sex, and center) to 0.49 (deattenuated correlation coefficient, taking into account the interindividual variability) for validity, according to the type of adjustment used. Corresponding values for retinol were 0.57 for reproducibility and 0.39–0.47 for validity.

In the control group, the major contributors to intake of each micronutrient considered were: for retinol, liver (64%), cheese (16%), and milk (6%); for beta-carotene equivalents, carrots (30%), green salads (10%), and spinach and Swiss chards (9%); for alpha-carotene, carrots (66%), mixed vegetable salads (14%), and vegetable soups (9%); for beta-carotene, carrots (26%), spinach and Swiss chards (15%), and peaches and apricots (9%); for beta-cryptoxanthin, citrus fruit (75%), peaches and apricots (12%), and mixed fruit juices (10%); for lycopene, tomatoes as a side dish (23%) and tomato-based foods, including pasta or rice with tomato sauce (43%), or “bolognese” sauce (17%); for lutein plus zeaxanthin, spinach and Swiss chards (36%), green salads (32%), and mixed vegetable salads (8%).

ORs of endometrial cancer and the corresponding 95% CIs were estimated through multiple logistic regression [19], using two different models: a basic model, conditioned on age and study center, and adjusted for year of interview; a multi-adjusted model, conditioned on age and study center, and adjusted for year of interview, education, body mass index (BMI, kg/m2), history of diabetes, age at menarche, menopausal status, parity, OC, and HRT use. To control for energy intake, the residuals of the linear regression of nutrients on energy intake were computed and included in the models instead of absolute intakes [20]. Quartiles were computed on the distribution of controls. Tests for trend for intake quartiles were based on the likelihood-ratio test between the models with and without a linear term for each micronutrient [19]. To test for interactions, the differences in −2log(likelihood) of the models with and without interaction terms were compared with the χ2 distribution with the same number of degrees of freedom as the interaction terms.

Results

Table 1 shows the distribution of 454 endometrial cancer cases and 908 controls according to age, study center, and other selected variables. By design, cases and controls had similar age and study center distributions. No clear association emerged with education. Cases reported BMI over 30 kg/m2, history of diabetes and of HRT use more frequently than controls.
Table 1

Distribution of 454 cases of endometrial cancer and 908 controls according to age, center, and other selected covariates. Italy, 1992–2006

Characteristic

Cases

Controls

No.

(%)

No.

(%)

Age (years)

    <40

19

(4.2)

38

(4.2)

    40–49

48

(10.5)

96

(10.5)

    50–59

140

(30.8)

280

(30.8)

    60–69

166

(36.6)

332

(36.6)

    ≥70

81

(17.8)

162

(17.8)

Study center

    Milan

140

(30.8)

280

(30.8)

    Naples

77

(17.0)

154

(17.0)

    Pordenone

237

(52.2)

474

(52.2)

Education (years)

    <7

263

(57.9)

553

(60.9)

    7–11

119

(26.2)

225

(24.8)

    ≥12

72

(15.9)

130

(14.3)

Body mass index (kg/m−2)a

    <25

131

(28.9)

420

(46.5)

    ≥25–<30

155

(34.1)

344

(38.0)

    ≥30

168

(37.0)

140

(15.5)

History of diabetes

    No

401

(88.3)

854

(94.0)

    Yes

53

(11.7)

54

(5.9)

Parity

    Nulliparous

68

(15.0)

126

(13.9)

    Parous

386

(85.0)

782

(86.1)

Menopausal statusa

    Pre/peri

83

(18.7)

174

(19.3)

    Post

360

(81.3)

726

(80.7)

Oral contraceptive use

    No

408

(89.9)

790

(87.0)

    Yes

46

(10.1)

118

(13.0)

Hormone replacement therapy use

    No

405

(89.2)

830

(91.4)

    Yes

49

(10.8)

78

(8.6)

aThe sum does not add up to the total because of missing values

Table 2 gives ORs of endometrial cancer and their 95% CI, estimated using different multivariate models, according to intake of retinol and selected carotenoids. The results of the two models were similar, though associations with beta-carotene and lutein plus zeaxanthin were slightly stronger using the multi-adjusted model. Comparing the highest to the lowest quartile of intake, significantly reduced ORs were found for beta-carotene (OR = 0.69, 95% CI, 0.48–0.99), beta-cryptoxanthin (OR = 0.65, 95% CI, 0.45–0.94), and lutein plus zeaxanthin (OR = 0.59, 95% CI, 0.41–0.85). Corresponding values were 1.31 (95% CI, 0.94–1.84) for retinol, 0.94 (95% CI, 0.66–1.34) for alpha-carotene, and 0.95 (95% CI, 0.68–1.34) for lycopene intake. For lutein plus zeaxanthin, the ORs were significantly below unity also in the second (OR = 0.58, 95% CI, 0.40–0.82) and third quartile of intake (OR = 0.60, 95% CI, 0.42–0.86). An inverse trend in risk emerged for beta-carotene equivalents (p = 0.04), beta-carotene (p = 0.04), beta-cryptoxanthin (p = 0.02), and lutein plus zeaxanthin (p < 0.01). Other models tested were: the multi-adjusted model, plus terms for tobacco smoking; the multi-adjusted model, plus terms for breastfeeding; the multi-adjusted model, plus terms for fruits and vegetables intake. None of these models materially changed the results. With reference to the latter, the ORs for the highest versus lowest quartile of intake were 1.28 (95% CI, 0.91–1.79) for retinol, 0.79 (95% CI, 0.53–1.16) for beta-carotene equivalents, 1.01 (95% CI, 0.70–1.45) for alpha-carotene, 0.74 (95% CI, 0.49–1.11) for beta-carotene, 0.61 (95% CI, 0.40–0.92) for beta-cryptoxanthin, 0.99 (95% CI, 0.71–1.40) for lycopene, and 0.61 (95% CI, 0.41–0.91) for lutein plus zeaxanthin.
Table 2

Odds ratios (OR) of endometrial cancer and corresponding 95% confidence intervals (CI) according to intake of retinol and selected carotenoids. Italy, 1992–2006

 

Quartile of intake, calculated on the residual valuesc

χ2, trend (p-value)

1st

2nd

3rd

4th

Retinol

    OR (95% CI)a

1d

0.79 (0.56–1.10)

0.80 (0.57–1.11)

1.22 (0.89–1.67)

1.73 (0.19)

    OR (95% CI)b

1d

0.76 (0.53–1.09)

0.86 (0.59–1.23)

1.31 (0.94–1.84)

3.28 (0.07)

Beta-carotene equivalents

    OR (95% CI)a

1d

0.92 (0.67–1.26)

0.74 (0.53–1.02)

0.76 (0.55–1.05)

3.93 (0.05)

    OR (95% CI)b

1d

0.91 (0.65–1.28)

0.71 (0.50–1.01)

0.74 (0.52–1.05)

4.16 (0.04)

Alpha-carotene

    OR (95% CI)a

1d

1.05 (0.76–1.44)

0.96 (0.69–1.32)

0.87 (0.63–1.21)

0.86 (0.35)

    OR (95% CI)b

1d

1.13 (0.80–1.58)

1.02 (0.72–1.45)

0.94 (0.66–1.34)

0.22 (0.64)

Beta-carotene

    OR (95% CI)a

1d

0.96 (0.70–1.32)

0.83 (0.60–1.16)

0.74 (0.53–1.04)

3.62 (0.06)

    OR (95% CI)b

1d

0.90 (0.64–1.26)

0.79 (0.55–1.13)

0.69 (0.48–0.99)

4.41 (0.04)

Beta-cryptoxanthin

    OR (95% CI)a

1d

0.93 (0.68–1.27)

0.80 (0.57–1.10)

0.62 (0.44–0.88)

8.02 (0.005)

    OR (95% CI)b

1d

1.00 (0.72–1.40)

0.86 (0.61–1.21)

0.65 (0.45–0.94)

5.82 (0.02)

Lycopene

    OR (95% CI)a

1d

0.75 (0.54–1.04)

0.79 (0.57–1.09)

0.94 (0.68–1.29)

0.12 (0.73)

    OR (95% CI)b

1d

0.73 (0.52–1.03)

0.88 (0.62–1.24)

0.95 (0.68–1.34)

0.01 (0.98)

Lutein plus zeaxanthin

    OR (95% CI)a

1d

0.61 (0.43–0.84)

0.64 (0.46–0.89)

0.70 (0.50–0.97)

4.01 (0.045)

    OR (95% CI)b

1d

0.58 (0.40–0.82)

0.60 (0.42–0.86)

0.59 (0.41–0.85)

7.37 (0.007)

aEstimates from multiple logistic regression, conditioned on age and study center, and adjusted for year of interview

bEstimates from multiple logistic regression, conditioned on age and study center, and adjusted for year of interview, education, body mass index, history of diabetes, age at menarche, menopausal status, parity, oral contraceptives use, and hormone replacement therapy use

cThe 25th, 50th, and 75th percentiles of absolute (non-residual) intake among controls were (μg/day): 170.8, 255.7, and 471.6 for retinol; 2963.7, 4028.3, and 5191.6 for beta-carotene equivalents; 388.9, 696.4, and 1102.9 for alpha-carotene; 3383.2, 4664.3, and 6021.2 for beta-carotene; 121.2, 284.8, and 480.5 for beta-cryptoxanthin; 4007.0, 5653.1, and 7684.2 for lycopene; 3106.1, 4609.5, and 6080.5 for lutein plus zeaxanthin

dReference category

Table 3 shows the ORs of endometrial cancer for the highest versus the lowest quartile of intake of selected carotenoids, in separate strata of relevant covariates. No significant heterogeneity emerged across strata of age, BMI, and tobacco smoking. With reference to strata of HRT use, no heterogeneity was observed for beta-carotene equivalents, beta-carotene, and lutein plus zeaxanthin, while the effect of beta-cryptoxanthin seemed restricted to non-HRT users (p-value for heterogeneity = 0.004). Alcohol consumption, fat intake, education, and menopausal status were also considered as potential effect modifiers, and no significant heterogeneity was found.
Table 3

Odds ratios (OR)a of endometrial cancer and corresponding 95% confidence intervals (CI) according to intake of selected carotenoids, in strata of relevant covariates. Italy, 1992–2006

 

Beta-carotene equivalents

Beta-carotene

Beta-cryptoxanthin

Lutein + zeaxanthin

Age group

    <60 years

    

    Cases:controls, 4th Q

54:114

52:112

45:112

59:115

    OR

0.76

0.81

0.57

0.59

    95% CI

0.45–1.29

0.47–1.40

0.33–0.96

0.34–1.04

    ≥60 years

    

    Cases:controls, 4th Q

47:113

43:115

39:115

50:112

    OR

0.65

0.54

0.66

0.54

    95% CI

0.40–1.07

0.33–0.90

0.39–1.11

0.33–0.90

p-value for heterogeneity

0.93

0.21

0.31

0.61

Body mass index

    <25 kg/m2

    

    Cases:controls, 4th Q

30:118

30:120

22:113

27:107

    OR

0.68

0.72

0.50

0.59

    95% CI

0.37–1.26

0.39–1.32

0.26–0.96

0.31–1.10

    ≥25 kg/m2

    

    Cases:controls, 4th Q

71:109

65:107

62:113

82:120

    OR

0.78

0.70

0.74

0.62

    95% CI

0.50–1.23

0.44–1.12

0.47–1.17

0.39–0.99

p-value for heterogeneity

0.65

0.77

0.28

0.87

Tobacco smoking

    Non-smokers

    

    Cases:controls, 4th Q

81:190

76:189

72:181

92:186

    OR

0.68

0.63

0.71

0.59

    95% CI

0.45–1.01

0.42–0.96

0.47–1.07

0.39–0.88

    Smokers

    

    Cases:controls, 4th Q

20:37

19:38

12:45

17:40

    OR

1.41

1.03

0.65

0.72

    95% CI

0.58–3.42

0.42–2.51

0.25–1.63

0.27–1.87

p-value for heterogeneity

0.37

0.63

0.37

0.09

HRT use

    Non-users

    

    Cases:controls, 4th Q

87:204

81:205

70:206

92:205

    OR

0.73

0.65

0.55

0.55

    95% CI

0.50–1.06

0.44–0.95

0.38–0.82

0.37–0.81

    Users

    

    Cases:controls, 4th Q

14:23

14:22

14:21

17:22

    OR

0.47

0.95

3.48

0.64

    95% CI

0.11–1.92

0.23–3.84

0.91–13.3

0.14–2.93

p-value for heterogeneity

0.66

0.74

0.004

0.54

aORs comparing the highest versus lowest quartile of intake. Estimates from multiple logistic regression, conditioned on age and study center, and adjusted for year of interview, education, body mass index, history of diabetes, age at menarche, menopausal status, parity, oral contraceptives use, hormone replacement therapy use, and total energy intake (residual method)

Discussion

Intake of beta-carotene, beta-cryptoxanthin, and lutein plus zeaxanthin was inversely associated with endometrial cancer risk in this Italian population. We found no effect of a few other carotenoids, including lycopene, and of retinol. Therefore, the protective effect seems to be independent from the provitamin A activity exercised by some carotenoids (i.e., alpha-carotene, beta-carotene, and beta-cryptoxanthin).

Possible mechanisms of action of carotenoids on cancer risk include their antioxidant functions against oxidative DNA damage [11], capability to induce gap junctional communication [21], induction of apoptosis in transformed cells [22], and regulation of immune response [23]. In vitro studies reported the capacity of beta-carotene, as an antioxidant, to quench oxygen-containing free radicals by several mechanisms, including addition of the radical to the carotenoid, hydrogen abstraction, and electron transfer [24]. Further, beta-carotene and cryptoxanthin have been found to stimulate gap junctional intercellular communication, and this property was independent from the antioxidant capacity of carotenoids [21]. Lutein might exert an anti-carcinogenic effect by its ability to interact with the mutagens 1-nitro pyrene and aflatoxin B1, or by stimulating genes involved in T-cell transformations activated by mitogens, cytokines, and antigens [25].

Our results add to the current literature on diet and endometrial cancer risk [26, 27]. With reference to beta-carotene, during the last decades several epidemiological studies indicated an inverse association with a few cancers [2830], but subsequent intervention trials denied any beneficial outcome of beta-carotene supplements [3134]. As in the majority of other observational investigations, in our study beta-carotene was inversely related to endometrial cancer risk [1, 36]. Interestingly, though there was a strong direct correlation between alpha- and beta-carotene (ρ = 0.83 between energy-adjusted intakes), an inverse association was found with beta-carotene alone. This is difficult to explain, as is the observation that adjustment for relevant covariates tended to strengthen the association between beta-carotene and endometrial cancer, while weakened that with alpha-carotene. Dietary sources of alpha- and beta-carotene are not, however, totally overlapping (for example, milk contains small quantities of beta-carotene, but no alpha-carotene).

To our knowledge, this is the only investigation reporting an inverse association between beta-cryptoxanthin and endometrial cancer risk. In fact, no relation emerged with cryptoxanthin in two earlier case–control studies [3, 7]. However, levels of intake differed substantially between this and previous studies, as the median intake was 285 μg/day in our population, and 67 and 106 μg/day in two North-American studies, respectively. This might at least in part be due to different food composition tables used in different studies. However, since in this investigation a reduced risk was evident in the highest quartile only, it is also possible that high levels of intake are needed to observe a protective effect. The significant effect modification by HRT use on the association between beta-cryptoxanthin and endometrial cancer might be due to chance, given the low number of HRT users.

Fruit was the main source of beta-cryptoxanthin in this population. One may therefore speculate that other correlated characteristics or components of fruit or, more in general, of the whole diet might explain the results obtained for beta-cryptoxanthin [35]. Though many of the carotenoids examined were highly correlated with each other, and our efforts to disentangle their effects were hampered by inherent problems of collinearity, all correlation coefficients between beta-cryptoxanthin and other carotenoids were below 0.3. Therefore, disentangling the effect of beta-cryptoxanthin was possible, and the inverse association remained similar after mutual adjustment for other carotenoids, as well as for total vegetable and fruit intake.

The strongest inverse association was with lutein plus zeaxanthin. However, there was no dose-risk effect, as the risk was decreased by approximately 40% in all quartiles of intake as compared to the lowest one. Thus, low intakes of lutein plus zeaxanthin appear to have a detrimental effect on endometrial cancer in this population.

Some previous studies found an inverse relation between lycopene and endometrial cancer [3, 6, 7]. In this population, where tomatoes and tomato sauces, the main sources of lycopene [36, 37], are frequently consumed, we found no relevant association between lycopene and endometrial cancer. Possible explanations of these inconsistent results include differences in the pattern of lycopene intake (i.e., according to amount, food sources, and associated dietary characteristics) between different populations.

In this study, as in most case–control investigations, some selection and recall bias are possible. The low refusal rate for cases and controls is reassuring against selection bias. However, in the absence of population-based cancer registries, we had no information on total number of cases. The main cause of loss of subjects was absence of the patient from the ward at time of the interviewer’s visit, while we had no physician refusal of interview. This is unlikely therefore to have introduced major selection bias. With reference to information bias, a recent cancer diagnosis might have influenced recall of diet for cases, although awareness of dietary hypotheses in endometrial cancer was unknown to the general population and the interviewers. Further, by interviewing women in the same hospital setting, the comparability of information between cases and controls is improved [38]. Another limitation might be that dietary habits of hospital controls may be different from those of the general population. However, we carefully excluded from the control group any women admitted for conditions associated with long-term modifications of diet. Major strengths of this study are its large size, combined to the collection of extensive dietary information using a satisfactorily reproducible and valid FFQ [17, 18], the comparable catchment areas of cases and controls, and the possibility of allowance for intake of energy and for several covariates in the analyses.

In conclusion, our results add quantitative data for a role of diet on endometrial cancer, and support a favorable effect of selected dietary carotenoids, independent of their provitamin A activity. Findings on these issues from epidemiological studies are not totally consistent, possibly reflecting different dietary patterns and levels of carotenoids intake in different populations. Further, there is the need to better understand the mechanisms of action of beta-carotene and other carotenoids on cancer risk, as this will allow us to clarify the inconsistent findings from different types of studies [24].

Acknowledgments

The authors thank Ms I. Garimoldi for editorial assistance.

Financial support

This work was conducted with contribution from the Italian Association for Cancer Research (AIRC), Italian League Against Cancer and Italian Ministry of Research (PRIN 2005). The work in this paper was undertaken while CLV was a Senior Fellow at the International Agency for Research on Cancer.

Copyright information

© Springer Science+Business Media B.V. 2008