Serum Fatty Acids and Risk of Cutaneous Melanoma: A Population-Based Case-Control Study

Background. Some observational studies have suggested that excess dietary intake of polyunsaturated fatty acids such as linoleic acid increases cutaneous melanoma risk. We aimed at examining the association between serum fatty acids and melanoma risk by conducting a population-based case-control study in a northern Italy community. Methods. The percentage composition of 12 fatty acids was determined in 51 newly diagnosed melanoma patients and 51 age- and sex-matched population controls by extracting total lipids from serum samples using thin layer and gas chromatography. Conditional logistic regression was used to estimate the relative risk of melanoma associated with tertiles of percentage composition of each fatty acid as well as groupings including saturated, monounsaturated, and polyunsaturated fatty acids. Results. We found a slightly increased melanoma risk for stearic and arachidic acids proportion, with and without adjustment for potential confounders. For an n-3 polyunsaturated fatty acid, docosapentaenoic acid, we found a male-specific direct association with melanoma risk. No other associations emerged for the other saturated, monounsaturated, and polyunsaturated fatty acids, individually or grouped by type. Conclusions. These findings do not suggest a major role of fatty acids, including linoleic acid, on risk of cutaneous melanoma, though their evaluation is limited by the small sample size.


Introduction
A steady increase of melanoma incidence rates has been observed in recent decades in many countries [1][2][3]. A report from England on trends in cancer occurrence in 1991-2000 ranks melanoma incidence as the most rapidly increasing among all site-speci�c neoplasms, and early detection due to screening does not appear to be the only explanation of the growing incidence of the disease [4].
is points to a role of environmental risk factors in increasing melanoma incidence, but the exact nature of these factors, apart from ultraviolet exposure, is still unclear [5]. In particular diet, a potentially modi�able risk factor, has long been suspected to play a role in melanoma occurrence, though the results of epidemiologic investigations carried out so far have yielded inconsistent results [6][7][8][9][10][11][12][13]. High intake of polyunsaturated fatty acids (PUFAs), and of linoleic acid in particular, have been linked to increased melanoma risk [6,10,14]. In contrast, high intake of fat [15] and of �sh rich in n-3 fatty acids [12,16] has been associated with reduced risk of the disease. Laboratory studies have reported toxic effects of fatty acids such as linoleic and arachidonic acid on distinct cell types such as leukemia cells [17], rabbit renal cells [18], human endothelial cells [19], and murine and human melanoma cell lines [20], with different potential mechanisms including in vivo isomerization into trans fatty acids [21]. In epidermal reconstructs, n-6 PUFAs administration resulted in increased epidermal oxidative damage, possibly leading to accumulated mutations and increased risk of skin cancer, especially melanoma [22].

Dermatology Research and Practice
Taking into account these �ndings and the scarcity of human studies, in the present investigation we explored the association between serum fatty acid distribution and the risk of cutaneous melanoma in a population-based case-control study.

Study Population.
We invited the patients newly diagnosed with cutaneous melanoma attending the Dermatologic Clinic of Modena University Hospital in the period 1999-2002 to participate in a population-based case-control study. Inclusion criteria were residence in Modena Province, an area in northern Italy with approximately 700,000 inhabitants, and a histologically con�rmed diagnosis of cutaneous melanoma without clinical evidence of metastasis. Fiy-nine of 84 eligible patients (72.0%) agreed to participate in the study. We randomly selected a resident of Modena Province, sex-and age-matched (±5 years) to each patient, from a database available at the Modena Local Health Authority. Eligible controls were contacted by telephone and enrolled in the study aer obtaining their informed consent. e participation rate among eligible population controls was 56.3%.

Data Collection.
We asked participants to go, early morning and fasting, to the Dermatologic Clinic of Modena where they provided a 10 cc blood sample and completed a questionnaire on demographic characteristics and lifestyle habits, including indicators of sun exposure and diet using the EPIC semiquantitative food frequency questionnaire [10,13]. Blood specimens were collected in EDTA tubes, immediately separated and stored at −20 ∘ C until analysis. Subjects whose blood samples were separated or stored more than 24 hours aer collection and whose samples were not available for other reasons were excluded from analyses, leaving 51 cases and their matched controls. A dermatologist determined each participant's phototype using the Fitzpatrick phototyping scale [22] and collected information about family history of melanoma, sun exposure, and other anamnestic characteristics.
e GC analysis was performed using an 8610 DANI instrument (Monza, Italy) equipped with a temperature vaporizer injector and a �ame ionization detector. We introduced the samples into the capillary column using a split-splitless technique, using a BPX70 column (30 m × 0.25 mm i.d.) with 0.25 m �lm thickness purchased from SGE (Ringwood, Australia). e gas conditions were: carrier gas helium at �ow rate of 0.8 mL/min; hydrogen 0.5 mL/min; make-up gas nitrogen 0.5 mL/min. e oven temperature increased from 130 ∘ C to 210 ∘ C at a rate of 3 ∘ C/min, from 210 ∘ C to 260 ∘ C at a rate of 5 ∘ C/min, then kept constant for 3 minutes. Identi�cation of FAME was obtained by comparison with the relative retention times of a commercial standard solution (Menhaden oil). e chromatograms were recorded and processed by computer-aided methods. e relative amount of each FAME (percent of total area) was quanti�ed by integrating the peak and dividing the results by total area for all fatty acids identi�ed. e detection limit was 0.01% for all fatty acids. Within-day coefficients of variation (CV) were based on the analysis of the same serum sample, extracted and analyzed 10 times during the same day. CV ranged from 0.6% for large peaks to 38.6% for small peaks. e between-day CVs were based on the analysis of one serum sample on separate days (up to 19 days). CV for all fatty acids were: 5.20% for palmitic, 28 on the total amount of fatty acids separated from the seric phospholipidic fraction. Determinations below the detection limit (0.01%) were assigned a value of 0.01%.

Statistical Analysis.
We computed odds ratios (OR) as an estimate of relative risk according to categories of fatty acid serum levels, based on a matched analysis using conditional logistic regression. For these analyses, levels of serum fatty acids were categorized into tertiles based on the distribution in controls, using the bottom tertile as the reference group. We also tested for linear trend by entering fatty acid composition as continuous variable. We repeated analyses adjusting for phototype, sun exposure, and education in multivariate analysis, and conducted sex-and age-(≥50 versus <50 years) strati�ed analyses. We also compared mean dietary intakes of selected fatty acids and fatty acids categories using Student's t-test.

Results
Demographic and clinical characteristics of cases and controls are reported in Table 1. Cases were similar to controls on educational attainment, had somewhat lower phototypes, and reported a higher occurrence of sunburns than controls. Table 2 summarizes the fatty acid distribution in sera of study participants, and Table 3 reports the ORs for cutaneous melanoma according to percentage composition of each serum fatty acid in unadjusted analyses. e proportions of stearic and arachidic acid were directly associated with increased melanoma risk, though these associations were limited and statistically imprecise. Compared with the lowest tertile, ORs for the highest tertile were 1.94 (95% CI, 0.61-6.20; P trend = 0.115) for stearic acid and 2.13 (95% CI, 0.61-7.46; P trend = 0.400) for arachidic acid. However, no association was detected for the overall SFAs group. No association was found with any of the other unsaturated fatty acids or with their groupings, nor with the ratios between the fatty acid categories.
We also compared cases and controls with respect to distribution of dietary intakes of the fatty acids which could be estimated with the EPIC food frequency questionnaire (saturated fatty acids, monounsaturated and polyunsaturated fatty acids, oleic, linoleic and linolenic acid). Mean intakes were similar in the two groups, with the exception of a somewhat higher linoleic acid intake in cases (8.48 g/day) compared with controls (7.75 g/day, P = 0.242).
ORs adjusted for educational attainment, skin phototype, and sun exposure are shown in Table 4. An increased risk of melanoma was con�rmed for stearic acid (OR for the highest tertile 3.00, 95%CI 0.56-16.07, P trend 0.101) and arachidic acid (OR for the highest tertile 2.05, 95% CI 0.45-9.40, P trend 0.334). A weak and statistically very unstable indication of an inverse association between melanoma risk and linoleic fatty acid also emerged (OR for the highest tertile 0.71, 95% CI 0.20-2.45, P trend 0.420). We performed a further multivariate analysis also adjusting for estimated dietary intakes of vitamin D, fruit and vegetables: none of these variables substantially modi�ed the results.
When we strati�ed on sex and age, there was little indication of any interaction with these factors. e only notable �nding was for the sex-speci�c analysis for docosapentaenoic acid, for which we observed a strong male-speci�c positive association with melanoma risk (OR 32.24 for the highest tertile, 95%CI 1.68-616.34, P trend 0.029), that was con�rmed in multivariate analysis.

Discussion
A direct relation between melanoma risk and dietary intake of fatty acids, particularly PUFAs, has been observed or suggested in some studies [6,10,14,23], though other epidemiologic investigations did not detect this association [24][25][26][27] or found an inverse association [16]. However, only one study on this topic using biomarkers (percentage of fatty acids in subcutaneous adipose tissue) has been performed to date [14], to the best of our knowledge, and therefore the present one appears to be the �rst study on fatty acids as a risk factor for melanoma analyzing serum distribution rather than assessing their dietary intakes from food questionnaires.
We found positive associations between melanoma incidence and serum proportions of stearic and arachidic acid. is �nding has some consistencies with the results of the case-control study by Millen et al. [9], who reported an increased melanoma risk for higher intakes of saturated fat. In contrast, our previous studies based on assessment of dietary intake did not suggest such association [10,13]. For PUFAs and linoleic acid, in the present investigation we observed no association and a weak inverse association with risk, respectively, in contrast with our previous observations in this study population of an increased melanoma risk associated with n-6 PUFAs levels and with linoleic acid itself [10,13], and with the excess melanoma risk among high consumers of cod liver oil, a supplement rich in n-3 PUFA, in a Norwegian cohort [6]. In sex-speci�c analyses, the n-3 PUFA docosapentaenoic acid was positively associated with melanoma risk in males, a result inconsistent with the observation by Veierod et al. of an increased melanoma risk associated with use of PUFA intake in females and not in males [6]. We speculate that the inconsistencies of the results between investigations based on dietary intake and the present study might be due to methodological issues such as the different indicator of fatty acid intake adopted (dietary intake of fatty acids instead of serum levels/percentage composition as in the present investigation), or by uncontrolled confounding. e inconsistencies may also be due to chance.
Strengths of the present study include its populationbased design, which reduces the risk of selection bias compared with hospital-based studies [28], and the adjustment for potential confounders, such as phototype, education, and sun exposure, known or suspected to in�uence melanoma risk. is study also has important limitations: its small sample size, which led to low statistical precision and hampered the detection of possible minor effects on risk; the possibility of residual confounding; and the use of percentage composition of serum fatty acids instead of their absolute levels, an approach which might have hidden some associations compared with an analysis based on serum concentrations of the fatty acids. In addition, the fatty acid assays were conducted postdiagnosis, and may not re�ect the relevant prediagnosis composition.
In conclusion, taken as a whole, this study yielded little evidence of an association between melanoma risk and circulating fatty acids. However, there was limited evidence for a direct relation between melanoma risk and stearic and arachidic acids, and docosapentaenoic acid in males, which may be worth further investigation.