Dietary fat intake and risk of skin cancer: A prospective study in Australian adults

Abstract

Although intakes of dietary fat have been associated with both basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) of the skin, the evidence is sparse and inconsistent. This study prospectively investigated the association between total dietary fat; saturated, polyunsaturated and monounsaturated fatty acids; and percent energy from fat in relation to BCC and SCC of the skin. At baseline in 1992, total fat intake and intake of fatty acids were assessed in an Australian community-based longitudinal study, using a validated semi-quantitative food frequency questionnaire in 1,057 adult residents (aged 25–75 years) in Nambour, Queensland. Information on demography, sun-sensitivity history and sun exposure factors were obtained using self-administered questionnaires. Associations with BCC and SCC in terms of persons newly affected and of tumor counts were assessed using Poisson and negative binomial regression models, respectively, based on incident, histologically-confirmed tumors occurring between 1992 and 2002. No significant linear trends were observed in overall risk of BCC or SCC of the skin with increasing total fat intake. However, in participants with a history of skin cancer, total fat intake (multivariable adjusted RR = 2.42, 95% CI = 1.20–4.88; p for trend = 0.01) was associated with increased numbers of SCC tumors comparing the highest to lowest tertile. In conclusion, SCC tumor risk increased as total fat intake increased in people with a history of skin cancer. Dietary fats were not associated with BCC occurrence. © 2009 UICC

It is well established that high exposure to ultraviolet (UV) radiation is the main environmental cause of the most common skin cancers, basal cell carcinoma (BCC) and squamous cell carcinoma (SCC).1 Nevertheless, there is increasing evidence that dietary factors play a role in UV-related skin carcinogenesis. Using animal models of SCC, Black and colleagues2 demonstrated that unsaturated fats when compared with equivalent levels of saturated fats enhanced photo-carcinogenesis with respect to tumor latency and multiplicity. It is postulated that the mechanism involved free radical reactions, specifically lipid peroxidation, and modulation of the immune system.3–5 High levels of dietary fat (omega 6 in particular) have been shown to significantly suppress specific immune responses in animal models and elevate levels of prostaglandin E2 (PGE2) which is known to act as an immunoregulator of T cell function and to modulate UV carcinogenesis.3

However, epidemiologic studies have failed to provide consistent evidence for a link between total dietary fat or any type of dietary fat and skin cancer occurrence. Although a 2-year, low-fat intervention regimen resulted in a significant reduction in the occurrence of BCC and SCC together in individuals who had had previous skin cancers,6, 7 the combined outcome made it impossible to assess the effect of the intervention on either BCC or SCC separately. No significant association between dietary fat intake and BCC incidence was found in a nested case–control study of participants from the EPIC-Norfolk cohort study8; nor in the Nurses Health Study9 after 4 years of follow-up. In the Physician's Health Follow-up Study,10 there was an inverse association between total fat intake and BCC incidence in participants without a history of skin cancer. Although a review of studies investigating dietary factors and BCC and SCC of the skin suggested a positive association between fat intake and both BCC and SCC of the skin,11 the 2007 WCRF report alluded to the fact that data were either of too low quality, too inconsistent or the number of studies too few to allow conclusions to be reached about the association between total fat and skin cancer.1

We have undertaken an investigation into the association between dietary fat and BCC and SCC of the skin, using data collected from 11 years prospective study of Australian adults from the community of Nambour Australia, taking full account of sun-exposure histories and other established risk factors for skin cancers. Based on the limited evidence available to date, we hypothesize that increased consumption of total dietary fats and fatty acids would be associated with increased risk of developing BCC and SCC of the skin.

BCC, basal cell carcinoma; BMI, body mass index; CI, confidence interval; FFQ, food frequency questionnaire; kJ, kilojoules; RRs, relative risks; SCC, squamous cell carcinoma; UV, ultraviolet; WFR, weighed food record.

Material and methods

The Nambour study participants

As part of the Nambour Skin Cancer Study, we conducted this 11-year prospective cohort study among adults who originally had been selected at random in 1986 from residents of Nambour, a sub-tropical community in Queensland, Australia.12 Of the original 2,095 adults aged 20–69 years in 1986,12 1,621 participated in a field trial to assess skin cancer prevention between 1992 and 1996.13, 14 Detailed descriptions of the community sample, the field trial and its outcomes have been reported previously12–16 but in summary, the Nambour Skin Cancer Prevention Trial was a randomized controlled trial with a 2 by 2 factorial design that evaluated the effectiveness of daily consumption of one 30 mg β-carotene tablet and daily application of sunscreen in preventing skin cancer. At baseline of this study (1992), 1,447 field trial participants completed a food-frequency questionnaire (FFQ) and standard questionnaires on complexion type, sun sensitivity, sun exposure and other personal characteristics.13, 14 This study was approved by the ethics committee of the Queensland Institute of Medical Research and all participants provided informed written consent.

Dietary assessment

A self-administered semi-quantitative food frequency questionnaire (FFQ), adapted from the US Nurses' Health Study FFQ developed by Willett and colleagues,17, 18 was used to assess food intake in 1992. Validation of this questionnaire in the Nambour study population has been reported elsewhere.19–22

Respondents recalled how often on average they consumed a given amount of a specific food item in the past 6 months for 129 food items, ranging from “never” to “4 or more times a day.” Additional questions on the FFQ collected information on trimming of visible fat from meat, frequency of fried foods eaten at home, fried take-away food and specific types of fats and oils used on bread, vegetables and in cooking. Average daily intake of total fat, saturated fat, mono-unsaturated and poly-unsaturated fat was calculated by expressing the frequency intake of each food item as a portion of daily use, which was then multiplied by the amounts (in grams) of the specified portion size and by the nutrient content of the food using Australian food composition table (NUTTAB95).23 The proportion of energy intake from total fat was also calculated.

BCC and SCC determination

Study participants were followed-up from trial baseline in February 1992 until the end of 2002 to ascertain all occurrences of BCC and SCC. Full-body skin examinations were conducted as part of the trial protocol by dermatologists in 1992, 1994 and 1996.14 All skin cancers diagnosed clinically were biopsied and confirmed histologically. In addition, participants carried a skin cancer treatment card on which the participants' doctors recorded all clinically diagnosed skin cancers at time of treatment. Every 3 months, participants were followed-up in study clinics and asked about any skin cancer treated since last contact. After 1996 when the intervention ceased, information on subsequent skin cancers was collected through questionnaires mailed twice-yearly to participants. In 2000, participants were offered a further full skin examination by a dermatologically trained physician, and suspected skin cancers were biopsied for histological confirmation.24 Finally, independent pathology laboratories throughout Queensland provided pathology reports for all skin cancers diagnosed among study participants in the entire follow-up period 1992–2002. These methods ensured virtually 100% ascertainment of histologically confirmed skin cancers in the study population.24 Standardized self-administered questionnaires completed by participants at baseline provided information on eye, hair and skin color, tanning ability of the skin, painful burns during life time, usual time spent outdoors (weekday, weekend), other personal characteristics including education, occupation, smoking habits and presence of selected medical conditions. In addition, trained personnel used standard procedures to measure height and weight of each participant. During a physical examination, degree of elastosis of the neck was recorded as an objective measure of long term sun exposure.

Statistical analysis

Dietary measurement error correction

In a previous FFQ validation study21 and in this study population, we showed that up to 20% of the variation in the difference between FFQ and weighed food record estimates of total fat and saturated fat intakes was explained by energy, sex, age, body mass index (BMI), occupation and medical condition. To correct the gram estimates of total dietary fat and fatty acids intake for these factors, we used the residual method described by Willett.18 Percent energy from total fat was not corrected for measurement error.

Two outcomes were used in the analysis: (i) incidence of persons affected (for person-based analysis) and (ii) incidence of tumor (for tumor-based analysis). Risk estimates of persons affected were derived by using generalized linear model specifying Poisson distribution with a robust error variance.25 Incidence of persons affected by a new BCC or SCC was calculated as number of persons diagnosed with BCC or SCC after the 1992 skin examination through to 31 December 2002, divided by the person-years of follow-up accumulated between these dates.

Risk estimates for tumor incidence were derived using generalized linear model with negative binomial distribution. Incidence of basal or squamous cell tumors was calculated as the number of new BCC or SCC diagnosed after the 1992 skin examination through to 31 December 2002, divided by the person-years of follow-up accumulated between these dates. Negative binomial distribution has been recommended for analyzing non-negative integer data with variance greater than the mean26 and provided the best fit for our tumor-count data. Person-years of follow-up were counted from baseline in 1992 until date of withdrawal from study, date of death or end of follow-up whichever came first. For further analyses, dietary intakes were categorized into 3 equal groups using tertile cut-points in the study population. Analyses were performed separately for BCC and SCC.

The multiple regression models adjusting for potential confounders were performed in steps: A basic model adjusted for the confounding effects of age (continuous); sex (male, female); trial treatment allocation groups (sunscreen treatment: yes, no; beta-carotene treatment: yes, no) and total energy (continuous) for BCC and SCC analyses. An expanded model adjusted for the additional effects of degree of solar elastosis of the neck (none, mild to moderate, severe); skin color (fair, medium, olive); skin cancer history before 1992 (yes, no) and smoking status in pack-years smoked (0, <7, 7–20, >20) for BCC analysis. The expanded model additionally adjusted for freckles on the back (yes, no) for SCC analysis. These confounders were selected based on their association with the exposure and outcome variables and also, based on previous studies of diet and skin cancer. Linear trends were assessed by assigning the numbers 1–3 to the lowest through highest intake groups, respectively, and modeling this as a continuous variable.

Stratification on the history of any skin cancer before 1992 was performed on the basis of a priori knowledge that people who have a history of skin cancer have an increased risk of developing subsequent skin cancers27–29 or other noncutaneous malignancies30, 31 and may, therefore, be more prone to risk modification by dietary factor. Previous analyses in our study population have shown this.32–35 Thus, we carried out a formal test of interaction between the dietary fats and history of any skin cancer by including a multiplicative interaction term in the linear models. We repeated the above analyses in 2 strata based on presence or absence of a history of any type of skin cancer if interactions of p values up to 0.1 were present. A positive skin cancer history was defined as the presence of any type of skin cancer before 1992. Information on skin cancer history before 1992 was based on skin cancers identified during skin examinations between 1986 when the cohort was established and 1992,12–16 and on self-reports of any type of skin cancer before 1986,12 with pathology verification of the subset of cancers reported between 1985 and 1986. All statistical analyses were performed in STATA software (version 9.0); all tests were 2-sided, and p values < 0.05 were considered significant.

Results

Of 1,447 study participants who completed the FFQ in 1992, 53 participants had 10% or more food consumption frequencies missing, while 34 participants reported energy intake levels outside the recommended normal ranges.18 These were excluded from the study, leaving 1,360 participants. Further correction for measurement error left a total of 1,057 participants (457males, 600 females) owing to missing values for 1 or more of the factors associated with measurement error.

The participants included in the present analyses were not significantly different from the original 1,621 field trial participants in terms of trial treatment allocation, age, sex, education, occupation type, smoking history, skin color, lifetime of sunburns and other measures of sun exposure (data not shown).

A total of 664 histologically confirmed BCC tumors were diagnosed in 267 participants and 235 histologically confirmed SCC tumors were diagnosed in 127 participants in the 11-year follow-up period. The overall crude incidence rates among participants without a prior history of skin cancer were 2,000 BCC tumors and 775 SCC tumors per 100,000 person-years while the corresponding rates among those with a prior history of skin cancer were more than 6 times higher at 13,936 BCC tumors and 4,744 SCC tumors per 100,000 person-years, respectively.

On average, participants in the highest tertile of total fat, saturated fatty acids and monounsaturated fatty acids were younger (p for trend < 0.01). Males were more likely to be high fat consumers compared to females, while participants with mild to moderate elastosis of the neck were more likely to consume polyunsaturated fat (p for trend = 0.03). Total energy intake increased significantly as amount of dietary fat consumed increased (p < 0.01) (Table I).

Table I. Selected Characteristics According to Tertiles of Measurement Error Corrected Dietary Total Fat and Fatty Acids, in 1,057 Participants of Nambour Skin Cancer Study 1992–2002

Variables	N	Total fat				Saturated fatty acid				Polyunsaturated fatty acid				Monounsaturated fatty acid
		Tertiles of intake			p for trend1	Tertiles of intake			p for trend1	Tertiles of intake			p for trend1	Tertiles of intake			p for trend1
		T1 Lowest	T2	T3 Highest		T1 Lowest	T2	T3 Highest		T1 Lowest	T2	T3 Highest		T1 Lowest	T2	T3 Highest
1 All statistical tests were 2-sided. 2 ANOVA was used for comparison of means. 3 Chi-square statistic for comparison of proportions.
Age (years) (Mean ± SD)2	1,057	52 ± 12	50 ± 13	48 ± 13	<0.01	51 ± 13	50 ± 13	48 ± 13	<0.01	49 ± 13	50 ± 13	50 ± 13	0.67	52 ± 12	50 ± 13	48 ± 13	<0.01
Gender (%)3					<0.01				<0.01				<0.01				<0.01
Female	600	71	59	42		70	58	43		66	57	47		71	60	40
Male	457	29	41	58		30	42	57		34	43	53		29	40	60
Smoking status in pack years (%)3
None	604	60	57	55	0.17	60	57	59	0.13	57	57	59	0.93	59	58	56	0.18
≤7 pack-years	155	14	14	15		15	14	14		15	14	14		14	14	15
>7–20 pack-years	125	8	13	14		10	12	12		10	12	12		8	13	14
>20 pack years	173	17	16	15		17	16	15		17	16	15		19	15	15
Sun exposure factors
Elastosis of the neck (%)3					0.37				0.17				0.04				0.26
None	209	18	24	19		17	24	20		23	21	16		18	25	19
Mild to moderate	511	49	45	50		52	45	48		46	45	55		49	44	51
Severe	335	32	31	31		31	31	32		31	34	29		32	31	30
Prior skin cancer history–present (%)3	288	31	27	24	0.10	30	28	23	0.11	26	27	26	0.92	31	26	25	0.21
Sun sensitivity factors
Skin color (%)3
Fair	593	57	55	55	0.41	55	57	54	0.40	59	56	52	0.16	58	56	53	0.21
Medium	359	37	39	37		38	35	40		35	40	38		35	38	40
Olive	65	6	6	8		7	8	6		6	7	8		7	6	7
Freckles on the back (%)3
Present	693	69	67	69	0.91	69	68	68	0.94	68	67	70	0.10	71	66	69	0.58
Total energy (MJ/day) mean ± SD2	1,057	6.5 ± 1.5	8.6 ± 1.2	11.5 ± 1.9	<0.01	6.6 ± 1.5	8.7 ± 1.5	11.3 ± 2.0	<0.01	7.0 ± 1.8	8.8 ± 1.9	10.8 ± 2.3	<0.01	6.6 ± 1.6	8.6 ± 1.3	11.4 ± 1.9	<0.01

No significant linear trends were observed between increased intakes of total fat and fatty acids overall, and risk of skin cancer based on analysis of persons affected (Table II). Similarly, no significant linear trends were observed between increased intakes of total fat and fatty acids overall, and risk of developing BCC or SCC tumors based on analysis of total incident tumors except for a suggestion of a nonsignificant linear increase in SCC tumor risk with increasing total fat intake (RR = 1.56; 95% CI = 0.91–2.68, p for trend = 0.09 after multivariable adjustment for potential confounders (Table III). Increased percent energy from total fat showed a positive association with SCC tumor risk but the trend was not significant (p for trend = 0.15).

Table II. Relative Risk (RR) and 95% Confidence Intervals (CI) of BCC and SCC Risk According to Tertiles of Dietary Total Fat and Fatty Acids Corrected for Measurement Error, and Percent Energy from Total Fat in Nambour Study Participants from 1992–2002 (Person-Based Analysis)

Dietary fats1	BCC				SCC
	Tertiles of intake			p for trend2	Tertiles of intake			p for trend2
	T1 (Lowest)	T2	T3 (Highest)		T1 (Lowest)	T2	T3 (Highest)
1 Dietary fats (except % energy from total fat) were corrected for energy, sex, age, body mass index (BMI), occupation and medical condition using residual method by Willett18. 2 All statistical tests were 2-sided. Relative risks (RRs) were obtained using a Generalized Linear Model (GLM) with Poisson distribution. Basic Model: adjusted for age (continuous), sex (male, female), and trial treatment allocation groups (sunscreen treatment: yes, no; beta-carotene treatment: yes, no) and total energy intake in kilojoules (continuous). 3 Expanded model: adjusted for the basic model and degree of elastosis of the neck (none, mild to moderate, severe); skin color (fair, medium, olive); history of skin cancer before 1992 (yes, no); and smoking status in pack-years smoked: (0, <7, 7–20, >20) for BCC analysis. 4 Expanded model: additional adjustment for freckles on the back (yes, no) for SCC analysis. 5 Not corrected for measurement error.
Saturated fatty acids
Median g (min, max)	20 (5–25)	30 (>25–36)	45 (>36–82)		20 (5–25)	30 (>25–36)	45 (>36–82)
Number of participants	353	352	352		353	352	352
Number of cases	73	62	82		35	29	37
Person-years of follow-up	3,550	3,553	3,570		3,550	3,553	3,570
Basic model (RR and 95% CI)	1.00 (reference)	0.92 (0.68–1.24)	1.09 (0.83–1.43)	0.54	1.00 (reference)	0.95 (0.61–1.48)	0.93 (0.61–1.41)	0.74
Expanded Model34 (RR and 95% CI)	1.00 (reference)	0.86 (0.65–1.15)	1.06 (0.82–1.39)	0.61	1.00 x(reference)	0.90 (0.58–1.40)	0.93 (0.62–1.40)	0.70
Monounsaturated fatty acids
Median g (min, max)	19 (7–23)	27 (>23–32)	38 (>32–77)		19 (7-23)	27 (>23–32)	38 (>32–77)
Number of participants	353	352	352		353	352	352
Number of cases	72	68	77		32	27	42
Person-years of follow-up	3,592	3,561	3,522		3,592	3,561	3,522
Basic model (RR and 95% CI)	1.00 (Referent)	0.98 (0.73-1.31)	1.05 (0.80-1.39)	0.72	1.00 (Referent)	0.88 (0.55–1.40)	1.20 (0.79–1.82)	0.38
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.92 (0.70–1.21)	0.98 (0.75–1.29)	0.91	1.00 (Referent)	0.87 (0.55–1.38)	1.22 (0.81–1.84)	0.32
Polyunsaturated fatty acids
Median g (min, max)	7 (2–9)	10 (>9–12)	14 (>12–35)		7 (2–9)	10 (>9–12)	14 (>12–35)
Number of participants	353	352	352		353	352	352
Number of cases	80	65	72		33	36	32
Person-years of follow-up	3,594	3,561	3,522		3,594	3,561	3,522
Basic model (RR and 95% CI)	1.00 (Referent)	0.82 (0.61–1.09)	0.93 (0.71–1.23)	0.60	1.00 (Referent)	1.08 (0.71–1.66)	1.06 (0.68–1.65)	0.78
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.80 (0.61–1.05)	0.91 (0.70–1.18)	0.43	1.00 (Referent)	1.14 (0.75–1.74)	1.20 (0.78–1.87)	0.39
Total fat
Median g (min, max)	53 (20–64)	75 (>64–86)	104 (>86–189)		53 (20–64)	75 (>64–86)	104 (>86–189)
Number of participants	353	352	352		353	352	352
Number of cases	74	68	75		30	30	41
Person-years of follow-up	3,585	3,538	3,554		3,585	3,538	3,554
Basic model (RR and 95% CI)	1.00 (Referent)	0.96 (0.72–1.29)	0.99 (0.75–1.32)	0.98	1.00 (Referent)	1.04 (0.65–1.65)	1.22 (0.80–1.88)	0.34
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.91 (0.69–1.21)	0.92 (0.70–1.21)	0.58	1.00 (Referent)	1.09 (0.69–1.72)	1.24 (0.82–1.89)	0.30
% Energy from total fat5
Median % (min, max)	28 (8–31)	33 (>31–35)	38 (>35–49)		28 (8–31)	33 (>31–35)	38 (>35–49)
Number of participants	454	453	453		454	453	453
Number of cases	95	102	70		43	49	25
Person-years of follow-up	4,209	4,409	4,329		4,209	4,409	4,329
Basic model (RR and 95% CI)	1.00 (reference)	1.15 (0.90–1.47)	0.87 (0.66–1.15)	0.38	1.00 (reference)	1.31 (0.90–1.90)	1.10 (0.73–1.67)	0.56
Expanded model34 (RR and 95% CI)	1.00 (reference)	1.16 (0.91–1.47)	0.88 (0.67–1.16)	0.40	1.00 (reference)	1.42 (0.98–2.06)	1.21 (0.80–1.82)	0.30

Table III. Relative Risk (RR) and 95% Confidence Intervals (CI) of BCC and SCC Tumour According to Tertiles of Total Dietary Fat and Fatty Acids Corrected for Measurement Error, and Percent Energy from Total Fat in Nambour Study Participants from 1992–2002 (Tumor-Based Analysis)

Dietary fats1	BCC				SCC
	Tertiles of intake			p for trend2	Tertiles of intake			p for trend2
	T1 (Lowest)	T2	T3 (Highest)		T1 (Lowest)	T2	T3 (Highest)
1 Dietary fats (except percent energy from total fat) were corrected for energy, sex, age, body mass index (BMI), occupation and medical condition using residual method by Willett18. 2 All statistical tests were 2-sided. Relative Risks (RRs) were obtained using a generalized linear model with negative binomial distribution. Basic model: adjusted for age, sex and trial treatment allocation groups (sunscreen treatment: yes, no; beta-carotene treatment: yes, no) and total energy intake in kilojoules (continuous). 3 Expanded Model: adjusted for the Basic model and degree of elastosis of the neck (none, mild to moderate, severe); skin colour (fair, medium, olive); history of skin cancer before 1992 (yes, no); and smoking status in pack-years smoked: (0, <7, 7-20, >20) for BCC analysis. 4 Expanded Model: additional adjustment for freckles on the back (yes, no) for SCC analysis. 5 Not corrected for measurement error.
Saturated fatty acids
Median g (min, max)	20 (5–25)	30 (>25–36)	45 (>36–82)		20 (5–25)	30 (>25–36)	45 (>36–82)
Number of participants	353	352	352		353	352	352
Number of tumors	183	117	203		53	45	82
Person-years of follow-up	3,550	3,553	3,570		3,550	3,553	3,570
Basic model (RR and 95% CI)	1.00 (reference)	0.68 (0.45–1.02)	1.02 (0.69–1.51)	0.14	1.00 (reference)	0.99 (0.56–1.76)	1.26 (0.74–2.15)	0.37
Expanded model34 (RR and 95% CI)	1.00 (reference)	0.65 (0.44–0.96)	1.09 (0.76–1.57)	0.56	1.00 (reference)	0.94 (0.53–1.69)	1.29 (0.76–2.20)	0.33
Monounsaturated fatty acids
Median g (min, max)	19 (7–23)	27 (>23–32)	38 (>32–77)		19 (7–23)	27 (>23–32)	38 (>32–77)
Number of participants	353	352	352		353	352	352
Number of tumors	169	132	202		51	42	87
Person-years of follow-up	3,592	3,561	3,522		3,592	3,561	3,522
Basic model (RR and 95% CI)	1.00 (Referent)	0.80 (0.53–1.20)	1.12 (0.76–1.66)	0.53	1.00 (Referent)	0.69 (0.38–1.23)	1.37 (0.81–2.33)	0.19
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.76 (0.52–1.13)	1.14 (0.79–1.64)	0.44	1.00 (Referent)	0.66 (0.39–1.20)	1.33 (0.78–2.27)	0.22
Polyunsaturated fatty acids
Median g (min, max)	7 (2–9)	10 (>9–12)	14 (>12–35)		7 (2–9)	10 (>9–12)	14 (>12–35)
Number of participants	353	352	352		353	352	352
Number of tumors	176	162	165		53	73	54
Person-years of follow-up	3,594	3,561	3,522		3,594	3,561	3,522
Basic model (RR and 95% CI)	1.00 (Referent)	0.85 (0.57–1.28)	1.04 (0.69–1.55)	0.86	1.00 (Referent)	1.48 (0.85–2.56)	1.21 (0.69–2.12)	0.53
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.92 (0.63–1.35)	1.18 (0.81–1.72)	0.39	1.00 (Referent)	1.57 (0.90–2.73)	1.25 (0.72–2.19)	0.45
Total fat
Median g (min, max)	53 (20–64)	75 (>64–86)	104 (>86–189)		53 (20–64)	75 (>64–86)	104 (>86–189)
Number of participants	353	352	352		353	352	352
Number of tumors	169	145	189		45	45	90
Person-years of follow-up	3,585	3,538	3,554		3,585	3,538	3,554
Basic model (RR and 95% CI)	1.00 (Referent)	0.88 (0.59–1.32)	1.05 (0.71–1.56)	0.78	1.00 (Referent)	0.99 (0.55–1.77)	1.59 (0.93–2.71)	0.07
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.89 (0.60–1.31)	1.05 (0.73–1.53)	0.76	1.00 (Referent)	1.04 (0.58–1.88)	1.56 (0.91–2.68)	0.09
% Energy from total fat5
Median % (min, max)	28 (8–31)	33 (>31–35)	38 (>35–49)		28 (8–31)	33 (>31–35)	38 (>35–49)
Number of participants	454	453	453		454	453	453
Number of tumors	249	211	204		68	101	66
Person-years of follow-up	4,209	4,409	4,329		4,209	4,409	4,329
Basic model (RR and 95% CI)	1.00 (reference)	0.92 (0.64–1.32)	0.98 (0.68–1.42)	0.93	1.00 (reference)	1.45 (0.90–2.35)	1.24 (0.74–2.07)	0.40
Expanded model34 (RR and 95% CI)	1.00 (reference)	1.09 (0.77–1.53)	1.06 (0.74–1.50)	0.74	1.00 (reference)	1.63 (1.01–2.62)	1.45 (0.87–2.44)	0.15

Formal testing for interaction between total fat intake and history of skin cancer suggested interaction of p = 0.09 in relation to SCC tumor risk.

Stratified analysis showed that among participants with a history of skin cancer (n = 278), there was a positive association between increased intake of total fat and SCC tumor risk (RR = 2.42, 95% CI = 1.20–4.88; p for trend = 0.01), after adjustment for sun exposure and sun sensitivity factors (Table IV). Overall results of multivariable analyses using uncorrected estimates of total fat and fatty acid intakes (n = 1,360) did not differ substantially from the error-adjusted estimates (data not shown).

Table IV. Relative Risk (RR) and 95% Confidence Intervals (CI) of SCC Tumour Risk According to Tertiles of Dietary Fats and Fatty Acids Corrected for Measurement Error, Stratified by History of Any Skin Cancer in Nambour Study Participants from 1992–2002 (Tumor-Based Analysis)

Dietary fat1	No skin cancer history (n = 744)				Skin cancer history present (n = 278)
	Tertiles of intake			p for trend2	Tertiles of intake			p for trend2
	T1 (Lowest)	T2	T3 (Highest)		T1 (Lowest)	T2	T3 (Highest)
1 Dietary fat was corrected for energy, sex, age, body mass index (BMI), occupation and medical condition using residual method by Willett18. 2 All statistical tests were 2-sided. Relative Risks (RRs) were obtained using a generalized linear model with negative binomial distribution. Basic model: adjusted for age, sex and trial treatment allocation groups (sunscreen treatment: yes, no; beta-carotene treatment: yes, no) and total energy intake in kilojoules (continuous). 3 Expanded model: adjusted for the basic model and degree of elastosis of the neck (none, mild to moderate, severe); skin color (fair, medium, olive); history of skin cancer before 1992 (yes, no); and smoking status in pack-years smoked: (0, <7, 7–20, >20) for BCC analysis. 4 Expanded model: additional adjustment for freckles on the back (yes, no) for SCC analysis. Interaction between total fat and history of skin cancer in relation to SCC risk was p = 0.09. 5 p value = 0.014 (highest versus lowest intake of total fat).
Total fat
Median g (min, max)	53 (19–64)	74 (>64–86)	104 (>86–189)		53 (15–64)	75 (>64–86)	106 (>86–161)
Number of participants	265	259	245		88	93	103
Number of tumors	19	16	21		26	29	69
Person-years of follow-up	2,726	2,605	2,494		859	932	1,069
Basic model (RR and 95% CI)	1.00 (Referent)	0.88 (0.36–2.15)	1.05 (0.44–2.54)	0.92	1.00 (Referent)	1.12 (0.53–2.38)	2.06 (1.05–4.06)	0.02
Expanded model34 (RR and 95% CI)	1.00 (Referent)	0.88 (0.35–2.21)	1.04 (0.44–2.49)	0.92	1.00 (Referent)	1.32 (0.61–2.88)	2.42 (1.20–4.88)5	0.01

Discussion

In this community-based prospective study of the association of dietary fats and skin cancer with 11 years follow-up, our findings do not support the hypothesis that increasing consumption of dietary fat increase BCC and SCC incidence overall. However, in our tumor based analysis, there was a suggestion of increased SCC tumor risk with increasing total fat intake. With further investigation, we found that increasing intake of total fat at a median intake of 106 g per day was significantly associated with a 2-fold increased SCC tumor risk in participants with a history of skin cancer after adjusting for sun sensitivity and sun exposure factors. In this study, the incidence of SCC tumors among those with prior history of skin cancer was 6 times greater compared to those without a skin cancer history. Thus, large differences in risk of SCC tumors exist between high and low susceptibility sub-groups in this high-risk population.

Our overall null finding between dietary fat intake and BCC incidence have been reasonably consistent with results from other cohort studies including the EPIC-Norfolk cohort study,8 and the Nurses' Health Study.36 An exception was the Health Professionals Follow-up study10 which showed an inverse association between total fat intake and BCC incidence, but the BCC diagnosis in that study was based on self-report. Results from an Australian cohort study with 56–80 months follow-up37 suggested a nonsignificant inverse association between dietary fat scores and BCC and SCC of the skin; however according to the authors, the dietary fat instrument used was imprecise as authors were unable to quantify dietary fat by type, and total energy intake was not taken into account. A dietary intervention study (<20% of energy from fat in the intervention arm, and 38% in the control), which suggested that a low-fat diet is protective against skin cancer, presented BCC and SCC as a combined outcome6 so it was not possible to know the relevance of the findings to the BCC alone and hence to the present study.

With respect to SCC, our overall null finding is reasonably consistent with results from a population-based case–control study in Southeast Arizona38 where no consistent patterns of association were observed between total fat, polyunsaturated fatty acid, total energy intake and SCC incidence.

Although we found no overall association between increased dietary fat intake and risk of BCC or SCC of the skin, our results for those with a history of skin cancer does support the body of literature,6, 32, 33 which shows that people with prior skin cancer do not benefit from a high fat diet. In previous studies in this population, we have shown a positive association between intake of unmodified or high fat dairy products (RR = 2.53, 95% CI: 1.15–5.54; p for trend = 0.03),32 dietary pattern characterized by high meat and fat intake (RR = 3.77; 95% CI: 1.65–8.63; p for trend = 0.002)33 and increased SCC tumor risk. In this study, respondents in the highest tertile consumed an average of 109 g total fat per day (median daily intake 104 g). This level of intake is 20% greater than the mean daily consumption of 91 g total fat (median 75 g) among persons 19 years and over for Australia, as estimated in the Australian 1995 National Nutrition Survey.39 Black and colleagues6 reported the only clinical trial where a decrease in the occurrence of nonmelanoma skin cancer occurred in patients with skin cancer history who were on a low fat diet (20% of energy from fat); however, our study could not be compared with that study because the authors combined their outcome (BCC and SCC). Nevertheless, Black and colleagues had demonstrated in experimental animal studies that both level of lipid intake and type can influence UV-induced carcinogenesis by reducing tumor latency period and increasing tumor multiplicity. From their study, it was shown that the point at which dietary lipid exert its effect is at the postinitiation stage.40 Our result of increased SCC tumor risk in participants with a skin cancer history agrees with that finding.

Animal experiments have consistently demonstrated that high fat intake may increase the sensitivity of the skin toward UV carcinogenesis; in humans, however, results have been inconsistent particularly for BCC. These conflicting findings show that the issue of fat intake and BCC risk in humans is apparently more complex and may involve several other factors including prior history of cancers, action of other nutrients including vitamins, minerals, antioxidants and the composition of dietary fats.

The strengths of our study include its population-based prospective design over a substantial follow-up period of 11 years, the availability of information on sun exposure and sun sensitivity factors in this study population, the effect of which we were able to take into account in our analyses. The study was based on histologically confirmed BCC and SCC tumors, which were ascertained through an extensive surveillance system comprising dermatologic examinations, questionnaires, doctors' records and independent reports from pathology laboratories. We thus consider misclassification of participants through misdiagnosis or missed diagnosis of these skin cancers unlikely. We acknowledge that dietary intake would inevitably be measured with error, and this undoubtedly would result in some misclassification, but we have avoided this by correcting our estimates for measurement errors, thereby minimizing such errors.

In conclusion, our findings show that high intakes of total fat increase the risk of developing SCC tumors among people who have a history of skin cancer.

Acknowledgements

We thank the Nambour skin cancer study participants for their continued interest and support. We acknowledge Professor Gail Williams of the Australian Centre for International and Tropical Health and Nutrition, University of Queensland and Ms. Nirmala Pandeya of the Queensland Institute of Medical Research, Queensland, Australia, for statistical advice.