Intake of Alcohol May Modify the Risk for Non-Melanoma Skin Cancer: Results of a Large Danish Prospective Cohort Study

ORIGINAL ARTICLE

Intake of Alcohol May Modify the Risk for Non-Melanoma Skin Cancer: Results of a Large Danish Prospective Cohort Study Allan Jensen1, Fatima Birch-Johansen1, Anne B. Olesen2, Jane Christensen1, Anne Tjønneland1 and Susanne K. Kjær1,3 Alcohol has not been linked definitively to non-melanoma skin cancer. We examined whether alcohol intake affects the risks for basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) using data on 54,766 persons enrolled in the prospective Diet, Cancer, and Health cohort. Statistical analyses were based on the Cox proportional hazards model. All hazard ratios (HRs) were multivariate adjusted. Adjustment for exposure to UVR was not possible, but all analyses were adjusted for factors related to susceptibility to UVR, including sun sensitivity, degree of freckling, and number of nevi. A total of 2,409 BCC cases and 198 SCC cases were diagnosed within a median follow-up of 11.4 years. Total current alcohol intake was not associated with BCC risk, but beverage-specific analyses showed an increased BCC risk associated with intake of wine (HR ¼ 1.05, 95% confidence interval (CI): 1.02–1.08, current average alcohol intake, per 10 g per day) and spirits (HR ¼ 1.11, 95% CI: 1.02–1.21) and a decreased risk with beer (HR ¼ 0.97, 95% CI: 0.93–1.00). No convincing associations were found between total alcohol intake and risk for SCC, perhaps because of the limited number of cases. Our findings indicate that alcohol intake may increase the risk for BCC, but the relations seemed to depend on beverage type. Journal of Investigative Dermatology (2012) 132, 2718–2726; doi:10.1038/jid.2012.198; published online 14 June 2012

INTRODUCTION The incidence of non-melanoma skin cancer (NMSC), comprising primarily basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), has increased greatly in white populations throughout the world (Madan et al., 2010). This increase may be due to behavioral changes resulting in increased exposure to UVR, which is the most important risk factor for NMSC (Young, 2009). Other well-established risk factors for NMSC include light skin color, a tendency to sunburn, light or red hair, and light eye color (Green et al., 2008). Modifiable lifestyle risk factors, such as diet, exposure to exogenous hormones, alcohol intake, and tobacco smoking, all of which have been linked to carcinogenesis, have not definitively been confirmed as risk factors for NMSC. 1

Virus, Lifestyle and Genes, Danish Cancer Society Research Center, Copenhagen O, Denmark; 2Department of Dermatology, Aarhus University Hospital, Aarhus Hospital, Aarhus C, Denmark and 3Gynecological Clinic, The Juliane Marie Center, Copenhagen University Hospital, University of Copenhagen, Copenhagen O, Denmark Correspondence: Allan Jensen, Virus, Lifestyle and Genes, Danish Cancer Society Research Center, Strandboulevarden 49, DK-2100 Copenhagen O, Denmark. E-mail: [email protected] Abbreviations: BCC, basal cell carcinoma; BMI, body mass index; CI, confidence interval; HR, hazard ratio; HRT, hormone replacement therapy; NMSC, non-melanoma skin cancer; SCC, squamous cell carcinoma Received 23 November 2011; revised 25 April 2012; accepted 30 April 2012; published online 14 June 2012

2718 Journal of Investigative Dermatology (2012), Volume 132

The few epidemiological studies on alcohol intake as a risk factor for NMSC have produced inconsistent results (Kune et al., 1992; Sahl et al., 1995; Corona et al., 2001; Fung et al., 2002; Freedman et al., 2003; Milan et al., 2003; Ansems et al., 2008). Three case–control studies and one cohort study found no association between current alcohol intake and risk for BCC (Sahl et al., 1995; Corona et al., 2001; Milan et al., 2003; Ansems et al., 2008), whereas two large cohort studies both found an increased risk for BCC associated with total current alcohol intake (Fung et al., 2002; Freedman et al., 2003). To our knowledge, the risk for SCC associated with alcohol intake has been addressed in only one study, in which no convincing associations were observed (Ansems et al., 2008). The available results are therefore not very consistent, and a link between alcohol and risk for NMSC has not been fully established. Most of the earlier studies were case–control studies, in which recall bias is a potential problem, especially with regard to alcohol intake. Furthermore, important potential confounders such as UV exposure, skin reaction to sun, and skin constitutional factors were not adjusted for in all the previous studies, perhaps contributing to some of the inconsistencies. To further address the association between alcohol intake and risks for BCC and SCC, we used data from the large prospective Diet, Cancer and Health cohort study (Tjonneland et al., 2007), which has detailed information on alcohol intake and several other lifestyle factors for 57,054 & 2012 The Society for Investigative Dermatology

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

Danish men and women. Denmark has one the lowest proportions of alcohol abstainers among western countries (World Health Organization, 2007) and may therefore be a good setting for studying the association between alcohol intake and risk for NMSC. RESULTS Among the 54,766 persons (26,411 men and 28,355 women) in the study cohort, 2,409 cases of BCC (1,185 cases in men and 1,224 cases in women), and 198 cases of SCC (118 cases in men and 80 cases in women) were diagnosed within a median follow-up period of 11.4 years. The mean age at cohort entry was 56 years for both genders. In all, 97% of persons were current alcohol users. For men, the highest alcohol intake was from beer (median, 6.7 g per day), which was markedly higher than that for women (median, 1.0 g per day). For women, the highest intake of alcohol was from wine (median, 5.4 g per day), which was comparable to the wine intake of men (median, 5.5 g per day). The median intake of spirits was low for both genders (men, 0.8 g per day; women, 0.3 g per day). Table 1 presents the characteristics of persons enrolled in the Diet, Cancer and Health study (1993–1997) according to total current average alcohol intake. In general, the distribution of potential confounders (skin reaction when exposed to strong sunlight, degree of freckling, number of nevi, body mass index (BMI), duration of education, hormone replacement therapy (HRT), and menopausal status) was not very different across levels of total current alcohol intake. Table 2 shows unadjusted hazard ratios (HRs) for the associations between potential confounders and BCC plus SCC. The risk for BCC increased markedly with increased skin sensitivity to sunlight, degree of freckling, number of nevi, duration of education (P-values for trend o0.001), and decreased BMI (HR ¼ 0.96, 95% confidence interval (CI): 0.94–0.97). Furthermore, current users of HRT had a higher risk for BCC than never users (HR ¼ 1.27, 95% CI: 1.12–1.45). Premenopausal women had a statistically nonsignificantly higher risk for BCC than postmenopausal women. The risk for SCC was markedly increased with increased skin sensitivity to sunlight and degree of freckling (P-values for trend o0.001) but not with number of nevi or duration of education. The associations between BMI, use of HRT, menopausal status, and risk for SCC were in general in the same direction as for BCC but were not statistically significant. Overall and gender-specific HRs according to total alcohol use are presented in Table 3. The fully adjusted HRs did not substantially differ from the unadjusted HRs or from HRs in which nevi and freckles were excluded as confounders (data not shown). Our results showed no association between alcohol status (‘‘other’’ including never, past, or occasional use vs current use) and risk for BCC. People who drank an average of 410–p30 or 430–p50 g alcohol per day (approximately one to five drinks a day) had statistically significantly higher HRs for BCC than the reference group (40–p10 g alcohol per day). In contrast, the risk for BCC was not significantly increased at the highest category of current alcohol intake (X50 g alcohol per day). When the linearity of the association was evaluated with a

linear spline, the within-category estimate of the dose–response relation between current alcohol intake and cancer risk showed no increased risk for BCC overall and in men but a statistically significant increased risk of 1.05 (95% CI: 1.01–1.09) for BCC in women associated with an increase in 10 g alcohol per day. No convincing association was observed between cumulative alcohol intake and risk for BCC. Current alcohol users had a significantly lower risk for SCC than other users (HR ¼ 0.51, 95% CI: 0.28–0.95). The overall decrease was due mainly to a statistically significant decreased risk in women (HR ¼ 0.38, 95% CI: 0.17–0.83) and to a lesser degree to a statistically nonsignificant decreased risk in men. No convincing association was found between current alcohol intake, cumulative alcohol intake, and risk for SCC. Table 4 presents overall and gender-specific HRs for BCC according to current beverage-specific alcohol use. Corresponding analyses were not performed for SCC because the small number of cases did not allow meaningful interpretation of the results. The adjusted HRs did not differ substantially from the unadjusted HRs or from HRs in which nevi and freckles were excluded as confounders (data not shown). People who drank an average of 410–p30 or 430–p50 g alcohol from wine per day had statistically significantly higher HRs for BCC than the reference group (40–p10 g alcohol per day), whereas the risk was not significantly increased at the highest category of wine intake (X50 g per day). Furthermore, we found that each extra intake of 10 g alcohol from wine per day among persons who drank wine was associated with an increase in risk of 1.05 (95% CI: 1.02–1.08). For men and women separately, similar statistically significant positive associations between wine intake and risk for BCC were observed. Similarly, a linear dose–response association between intake of alcohol from spirits and risk for BCC was observed (HR ¼ 1.11, 95% CI: 1.02–1.21), whereas each extra intake of 10 g of alcohol from beer per day decreased the overall risk for BCC (HR ¼ 0.97, 95% CI: 0.93–1.00). Accordingly, the linear dose–response beverage-specific associations were significantly different (Po0.001 for heterogeneity). As both beer and spirits were used much more by men than women, the overall results primarily reflect the associations between beer intake and risk for BCC among men. We also investigated whether the risk for BCC associated with alcohol intake (measured as drinking frequency, where frequent drinkers consumed alcohol X5 times a week and less frequent drinkers consumed alcohol o5 times per week) varied according to use of HRT, skin reaction, and skin characteristics (degree of freckling and number of nevi), i.e., whether there are any interactions between these potential risk factors for BCC. The risk for BCC associated with alcohol intake was not significantly affected by any of these potential risk factors, as none of the interaction terms was statistically significant (all P-values 40.05; data not shown). DISCUSSION The results of this large prospective cohort study indicate that the risk for BCC increases with current alcohol intake; however, the association differed by type of beverage, intake www.jidonline.org 2719

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

Table 1. Characteristics of persons (n=54,766) enrolled in the Danish ‘‘Diet, Cancer and Health’’ study (1993–2007) according to current average alcohol intake Current average alcohol intake, g per day Characteristic 2

1

0

Occasional

40–p10

410–p30

430–p50

450

979 (2%)

775 (1%)

19,529 (36%)

19,018 (35%)

9,433 (17%)

5,032 (9%)

No. of men (%)2

418 (2%)

287 (1%)

5,783 (22%)

10,101 (38%)

5,658 (21%)

4,164 (16%)

No. of women (%)2

561 (2%)

488 (2%)

13,746 (48%)

8,917 (31%)

3,775 (13%)

868 (3%)

No. of persons (%)

No. of persons diagnosed with BCC (%)2

30 (1%)

25 (1%)

789 (33%)

884 (37%)

485 (20%)

196 (8%)

No. of men diagnosed with BCC (%)2

14 (1%)

8 (1%)

231 (19%)

490 (41%)

287 (24%)

155 (13%)

No. of women diagnosed with BCC (%)2

16 (1%)

17 (1%)

558 (46%)

394 (32%)

198 (16%)

41 (3%)

No. of persons diagnosed with SCC (%)2

5 (3%)

6 (3%)

53 (27%)

76 (38%)

39 (20%)

19 (10%)

No. of men diagnosed with SCC (%)2

2 (2%)

2 (2%)

24 (20%)

45 (38%)

27 (23%)

18 (15%)

No. of women diagnosed with SCC (%)2

3 (4%)

4 (5%)

29 (36%)

31 (39%)

12 (15%)

Mean age at entry (years±SD)

57.3±4.5

57.3±4.5

56.9±4.4

56.6±4.4

56.4±4.3

56.3±4.3

Mean BMI (kg m–2±SD)

26.3±4.9

26.8±5.3

26.3±4.4

25.8±3.7

25.7±3.7

26.6±3.9

270 (6%)

1 (1%)

Skin reaction when exposed to strong sunlight (%) Redness, pain, and blistering

86 (9%)

77 (10%)

1,437 (7%)

1,106 (6%)

507 (6%)

Redness, pain, and peeling

129 (13%)

113 (15%)

3,143 (16%)

3,015 (16%)

1,391 (15%)

686 (14%)

Redness, then tan

490 (50%)

359 (46%)

10,903 (56%)

11,101 (58%)

5,560 (59%)

2,830 (56%)

Only tan

274 (28%)

226 (29%)

4,046 (21%)

3,796 (20%)

1,975 (21%)

1,246 (25%)

Number of nevi (%) None

398 (41%)

316 (41%)

5,903 (30%)

6,575 (35%)

3,525 (37%)

2,382 (47%)

Few

309 (32%)

247 (32%)

6,716 (34%)

6,728 (36%)

3,418 (36%)

1,639 (33%)

Moderate

205 (21%)

168 (22%)

5,270 (27%)

4,363 (23%)

1,912 (20%)

769 (15%)

67 (7%)

44 (6%)

1,640 (8%)

1,352 (7%)

578 (6%)

242 (5%)

Many

Degree of freckling (%) None

401 (41%)

316 (41%)

6,843 (35%)

6,871 (36%)

3,539 (38%)

2,239 (45%)

Few

432 (44%)

344 (44%)

9,297 (48%)

9,284 (49%)

4,518 (48%)

2,264 (45%)

Moderate

125 (13%)

104 (13%)

2,984 (15%)

2,540 (13%)

1,229 (13%)

482 (10%)

21 (2%)

11 (1%)

405 (2%)

323 (2%)

147 (2%)

47 (1%)

Short (o8 years)

413 (42%)

384 (50%)

7,516 (38%)

5,580 (29%)

2,477 (26%)

Medium (8–10 years)

409 (42%)

312 (40%)

8,953 (46%)

9,153 (48%)

4,293 (46%)

2,135 (42%)

Long (410 years)

157 (16%)

79 (10%)

3,060 (16%)

4,285 (23%)

2,663 (28%)

1,244 (25%)

Never use

329 (58%)

274 (56%)

7,844 (57%)

4,851 (54%)

1,977 (52%)

447 (51%)

Past use

100 (18%)

85 (17%)

2,190 (16%)

1,387 (16%)

563 (15%)

137 (16%)

Current use

132 (24%)

129 (27%)

3,712 (27%)

2,679 (30%)

1,235 (33%)

284 (33%)

Postmenopausal

507 (90%)

433 (88%)

11,601 (84%)

7,312 (82%)

3,026 (80%)

739 (85%)

Premenopausal

54 (10%)

58 (12%)

2,145 (16%)

1,605 (18%)

749 (20%)

129 (15%)

Many

Duration of education (%) 1,653 (33%)

Hormone replacement therapy (%)3

Menopausal status (%)3

Abbreviations: BCC, basal cell carcinoma; BMI, body mass index; SCC, squamous cell carcinoma. 1 Occasional drinkers were defined as people who reported either no intake of alcohol on the food frequency questionnaire but some drinking occasions on the lifestyle questionnaire, or reported no drinking occasions on the lifestyle questionnaire but some intake of alcohol on the food-frequency questionnaire. 2 Row percentages. All other percentages in the table are column percentages. 3 Women only.

2720 Journal of Investigative Dermatology (2012), Volume 132

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

Table 2. Overall (both genders combined, n=54,766) unadjusted HRs for BCC and SCC according to potential confounders at baseline BCC Potential confounders

Cases/entire cohort

SCC Unadjusted HR (95% CI)

Cases/entire cohort

Unadjusted HR (95% CI)

Skin reaction when exposed to strong sunlight Only tan

373/11,563

1.00

30/11,563

Redness, then tan Redness, pain, and peeling Redness, pain, and blistering

1.00

1,419/31,243

1.44 (1.29–1.62)

119/31,243

1.59 (1.07–2.38)

462/8,477

1.78 (1.55–2.04)

35/8,477

1.79 (1.10–2.92)

155/3,483

1.42 (1.18–1.71)

14/3,483

1.71 (0.90–3.22)

o0.001

P-value for trend

o0.05

Number of nevi None Few

747/20,209

1.00

85/20,209

1.00

1,217/26,139

1.31 (1.19–1.44)

89/26,139

0.92 (0.68–1.24)

Moderate

382/7,464

1.49 (1.31–1.69)

20.7464

0.80 (0.49–1.32)

Many

63/954

1.93 (1.49–2.49)

4/954

1.26 (0.46–3.45)

o0.001

P-value for trend

0.54

Degree of freckling None

643/19,099

1.00

63/19,099

1.00

Few

849/19,057

1.39 (1.25–1.54)

70/19,057

1.26 (0.89–1.78)

1.76 (1.58–1.97)

51/12,687

1.54 (1.05–2.26)

1.85 (1.58–2.16)

14/3,923

1.37 (0.76–2.48)

Moderate Many

694/12,687 223/3,923

P-value for trend

o0.001

o0.05

Per unit BMI (kg m–2)

0.96 (0.94–0.97)

0.97 (0.93–1.01)

Duration of education (%) Short (o8 years)

699/18,023

1.00

71/18,023

1.00

Medium (8–10 years)

1,126/25,255

1.22 (1.11–1.34)

81/25,255

0.96 (0.70–1.34)

Long (410 years)

584/11,488

1.40 (1.25–1.56)

46/11,488

1.17 (0.80–1.69)

o0.001

P-value for trend

0.50

Hormone replacement therapy (%)1 Never use

606/15,143

1.00

42/15,143

1.00

Past use

175/4,149

1.03 (0.87–1.22)

Current use

394/7,884

1.27 (1.12–1.45)

23/7,884

11/4,149

0.83 (0.43–1.62) 1.05 (0.63–1.75)

1.00

69/23,615

1.00

1.15 (0.97–1.37)

11/4,740

1.46 (0.70–3.05)

Menopausal status (%)1 Postmenopausal

1,018/23,615

Premenopausal

206/4,740

Abbreviations: BCC, basal cell carcinoma; BMI, body mass index; CI, confidence interval; HR, hazard ratio; SCC, squamous cell carcinoma. 1 Women only.

of wine, and spirits increasing the risk and a non-monotonic association between beer intake and BCC. Furthermore, no association between cumulative alcohol intake and risk for BCC was observed. Our results, therefore, indicate that

current alcohol intake is a more important determinant of this cancer than earlier lifetime exposure. No convincing associations between total alcohol intake and risk for SCC were observed. www.jidonline.org 2721

Current

2722 Journal of Investigative Dermatology (2012), Volume 132

1.11 (1.00–1.22)

1.26 (1.12–1.41)

1.03 (0.88–1.21)

884/19,018

485/9,433

196/5,032

410–p30

430–p50

450

1.15 (1.04–1.28)

1.02 (0.86–1.21)

0.83 (0.66–1.04)

584/11,575

159/3,591

81/2,413

450–p100

4100–p150

4150

72/2,265

142/3,184

431/8,571

523/12,003

17/388

155/4,164

287/5,658

490/10,101

231/5,783

14/418

1,163/25,706

22/705

Cases/entire cohort

1.00

Adjusted HR (95% CI)2

1.00 (0.97–1.02)

0.85 (0.66–1.09)

1.08 (0.90–1.30)

1.19 (1.05–1.36)

1.00

1.03 (0.63–1.67)

1.01 (0.99–1.04)

1.09 (0.89–1.34)

1.29 (1.09–1.54)

1.18 (1.01–1.38)

1.00

0.96 (0.56–1.66)

1.24 (0.81–1.88)

Men

BCC

9/148

17/407

153/3,004

949/22,147

96/2,649

41/868

198/3,775

394/8,917

558/13,746

16/561

1,191/27,306

33/1,049

Cases/entire cohort

1.00

Adjusted HR (95% CI)3

1.04 (0.99–1.10)

1.47 (0.76–2.85)

0.99 (0.61–1.61)

1.15 (0.97–1.37)

1.00

0.93 (0.75–1.15)

1.05 (1.01–1.09)

1.22 (0.89–1.68)

1.26 (1.07–1.49)

1.04 (0.91–1.19)

1.00

0.79 (0.48–1.29)

1.19 (0.84–1.69)

Women

13/2,413

12/3,591

52/11,575

112/34,150

9/3,037

19/5,032

39/9,433

76/19,018

53/19,529

5/979

187/53,012

11/1,754

Cases/entire cohort

1.02 (0.96–1.09)

1.25 (0.69–2.27)

0.75 (0.41–1.39)

1.08 (0.77–1.53)

1.00

0.91 (0.46–1.81)

1.03 (0.97–1.10)

1.25 (0.72–2.14)

1.41 (0.93–2.16)

1.36 (0.96–1.94)

1.00

1.87 (0.75–4.68)

0.51 (0.28–0.95)

1.00

Adjusted HR (95% CI)1

Combined

12/2,265

11/3,184

43/8,571

52/12,003

0/388

18/4,164

27/5,658

45/10,101

24/5,783

2/418

114/25,706

4/705

Cases/entire cohort

1.02 (0.95–1.10)

1.26 (0.67–2.38)

0.79 (0.41–1.52)

1.15 (0.77–1.73)

1.00

—

1.03 (0.96–1.11)

1.23 (0.66–2.28)

1.24 (0.71–2.16)

1.10 (0.67–1.80)

1.00

1.27 (0.30–5.40)

0.70 (0.26–1.89)

1.00

Adjusted HR (95% CI)2

SCC Men

7/1,049

1/148

1/407

9/3,004

60/22,147

9/2,649

1/868

12/3,775

31/8,917

29/13,746

3/561

73/27,306

1.08 (0.90–1.30)

2.60 (0.36–19.01)

0.79 (0.11–5.76)

1.04 (0.51–2.11)

1.00

1.20 (0.59–2.46)

1.05 (0.90–1.21)

0.56 (0.08–4.12)

1.56 (0.78–3.09)

1.66 (1.00–2.77)

1.00

2.68 (0.81–8.83)

0.38 (0.17–0.83)

1.00

Adjusted HR (95% CI)3

Women Cases/entire cohort

Abbreviations: BCC, basal cell carcinoma; CI, confidence interval; HR, hazard ratio; SCC, squamous cell carcinoma. 1 Stratified by gender and adjusted for age, sun sensitivity (redness, pain, and blistering; redness, pain, and peeling; redness then tan; or only tan), degree of freckling (none, few, moderate, or many), number of nevi (none, few, moderate, or many), duration of school education (short, medium, or long), and body mass index (continuous variable). 2 Adjusted for age, sun sensitivity (redness, pain, and blistering; redness, pain, and peeling; redness then tan; or only tan), degree of freckling (none, few, moderate, or many), number of nevi (none, few, moderate, or many), duration of school education (short, medium, or long), and body mass index (continuous variable). 3 Adjusted for age, sun sensitivity (redness, pain, and blistering; redness, pain, and peeling; redness then tan; or only tan), degree of freckling (none, few, moderate, or many), number of nevi (none, few, moderate, or many), duration of school education (short, medium, or long), body mass index (continuous variable), menopausal status (pre- or post-menopausal); and use of hormone replacement therapy at baseline (never, past, or current). 4 Includes ‘‘never’’, ‘‘past’’, and ‘‘occasional’’ drinkers. 5 Includes ‘‘never’’ and ‘‘past’’ drinkers. 6 Total intake from age of 20 years to 1 year before baseline, calculated in ‘‘drink-years’’ (1 drink per day in 1 year).

1.00 (0.97–1.02)

1.00

Cumulative intake, per 25 ‘‘drink-years’’

0.93 (0.77–1.13)

113/3,037

1,472/34,150

40–p50

0

Cumulative intake (‘‘drink-years’’)6

1.01 (0.99–1.04)

1.00

Current average alcohol intake, per 10 g per day

0.83 (0.58–1.20)

30/979

789/19,529

1.24 (0.95–1.62)

1.00

Adjusted HR (95% CI)1

40–p10

05

Current average alcohol intake (g per day)

55/1,754

2,354/53,012

Other4

Alcohol status

Alcohol intake

Cases/entire cohort

Combined

Table 3. HRs for BCC and SCC among all study subjects combined (n=54,766), men (n=26,411), and women (n=28,355) according to total alcohol intake status, total current average alcohol intake, and total cumulative alcohol intake

A Jensen et al.

Alcohol and Risk for Non-Melanoma Skin Cancer

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

Table 4. HRs for BCC among all study subjects combined (n=54,766), men (n=26,411), and women (n=28,355) according to current average beverage-specific alcohol intake Combined

Current alcohol intake

Cases/entire cohort1

Men

Adjusted HR (95% CI)2

Cases/entire cohort1

Women Adjusted HR (95% CI)3

Cases/entire cohort1

Adjusted HR (95% CI)4

Wine (g alcohol per day) 05

73/2,197

40 to p10

1,574/37,839

410 to p30 430 to p50 450

54/1,327

1.06 (0.80–1.41)

35/1,070

0.97 (0.63–1.51)

38/1,127

1.19 (0.81–1.75)

1.00

744/17,641

1.00

830/20,198

1.00

307/6,027

1.14 (1.01–1.29)

161/2,977

1.16 (0.98–1.38)

146/3,050

1.09 (0.91–1.30)

401/7,364

1.25 (1.12–1.40)

211/3,912

1.20 (1.02–1.40)

190/3,464

1.28 (1.09–1.51)

0.98 (0.74–1.29)

34/811

1.04 (0.73–1.47)

20/516

Current average intake, per 10 g alcohol per day

1.05 (1.02–1.08)

1.04 (1.00–1.08)

0.98 (0.62–1.53) 1.06 (1.00–1.10)

Beer (g alcohol per day) 05

61/6,607

0.97 (0.84–1.12)

40/1,231

0.74 (0.51–1.08)

210/5,376 934/21,160

1.02 (0.87–1.19)

40 to p10

1,626/36,323

1.00

692/15,163

1.00

410 to p30

332/5,996

1.18 (1.05–1.34)

270/4,697

1.24 (1.07–1.43)

62/1,299

430 to p50

148/3,801

0.90 (0.76–1.08)

133/3,372

0.96 (0.80–1.16)

15/429

0.89 (0.53–1.49)

450

53/2,039

0.70 (0.53–0.93)

50/1,948

0.75 (0.56–1.01)

3/91

0.91 (0.29–2.83)

Current average intake, per 10 g alcohol per day

0.97 (0.93–1.00)

0.97 (0.94–1.01)

1.00 1.09 (0.84–1.41)

1.03 (0.94–1.12)

Spirits (g alcohol per day) 05

272/7,686

40 to p10

2,095/23,292

410 to p30

36/772

1.06 (0.76–1.48)

27/571

1.15 (0.78–1.69)

9/201

1.00 (0.52–1.93)

430 to p50

4/90

1.16 (0.44–3.11)

3/73

1.16 (0.37–3.63)

1/17

1.68 (0.24–12.00)

450

2/32

1.95 (0.49–7.82)

1/25

1.77 (0.25–12.66)

1/7

3.09 (0.43–22.09)

Current average intake, per 10 g alcohol per day

0.87 (0.76–1.00) 1.00

91/2,450 1,063/23,292

1.11 (1.02–1.21)

1.00 (0.78–1.27) 1.00

1.16 (1.05–1.29)

181/5,236 1,032/22,894

0.83 (0.70–0.99) 1.00

1.04 (0.88–1.23)

Abbreviations: BCC, basal cell carcinoma; CI, confidence interval; HR, hazard ratio. Total number of cases/entire cohort in each beverage-specific category (wine, beer, and spirits). 2 Stratified by gender and adjusted for age, sun sensitivity (redness, pain, and blistering; redness, pain, and peeling; redness then tan; or only tan), degree of freckling (none, few, moderate, or many), number of nevi (none, few, moderate, or many), duration of school education (short, medium, or long), body mass index (continuous variable) and mutually adjusted for the various types of alcohol. 3 Adjusted for age, sun sensitivity (redness, pain, and blistering; redness, pain, and peeling; redness then tan; or only tan), degree of freckling (none, few, moderate, or many), number of nevi (none, few, moderate, or many), duration of school education (short, medium, or long), body mass index (continuous variable) and mutually adjusted for the various types of alcohol. 4 Adjusted for age, sun sensitivity (redness, pain, and blistering; redness, pain, and peeling; redness then tan; or only tan), degree of freckling (none, few, moderate, or many), number of nevi (none, few, moderate, or many), duration of school education (short, medium, or long), body mass index (continuous variable), menopausal status (pre-menopausal or post-menopausal), use of hormone replacement therapy at baseline (never, past, or current use) and mutually adjusted for the various types of alcohol. 5 No intake of the specific type of beverage. 1

The association between alcohol intake and risk for BCC was investigated in six previous studies (Sahl et al., 1995; Corona et al., 2001; Fung et al., 2002; Freedman et al., 2003; Milan et al., 2003; Ansems et al., 2008), our results being in line with those of the two large cohort studies (Fung et al., 2002; Freedman et al., 2003). In the largest cohort study to date, with 107,975 people, Fung et al. (2002) identified 6,088 cases of BCC and reported a significantly increased risk associated with total alcohol intake for both genders. Alcohol

intake from beer was not associated with the risk for BCC in either gender, but the risk was significantly associated with increasing daily intake of spirits for men but not for women. Furthermore, Fung et al. (2002) found an increased risk for BCC associated with intake of white wine in men but a decreased risk associated with intake of red wine in women. Freedman et al. (2003), using data from a cohort study of 68,371 people, identified 1,360 cases of BCC and found an increasing risk with increasing overall alcohol intake; they www.jidonline.org 2723

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

did not, however, differentiate between the types of alcohol beverage. In contrast to our results, four studies found no association between alcohol intake and risk for BCC (Sahl et al., 1995; Corona et al., 2001; Milan et al., 2003; Ansems et al., 2008). Most of these studies were, however, case–control studies, which may have problems related to recall bias (Sahl et al., 1995; Corona et al., 2001; Milan et al., 2003), and two studies lacked adjustment for UV exposure and skin constitutional factors (Sahl et al., 1995; Milan et al., 2003). Several biological mechanisms may explain the increased risk for BCC found in our study. Alcohol may promote NMSC by impairing cell-mediated and humoral immunological functions, by interfering with normal DNA replication and mitosis in the skin, and by alterating oncogenes and tumorsuppressor genes (e.g., mutations in the p53 gene) (Merimsky and Inbar, 1999). Additionally, acetaldehyde, a by-product of ethanol, is known to have direct mutagenic or carcinogenic effects (Poschl and Seitz, 2004). It may also induce skin inflammation, which may interfere with keratinocyte DNA synthesis and repair, which can promote skin carcinogenesis (Coutelle et al., 2004). Finally, alcohol may act as a cocarcinogen by a photosensitizing effect, enhancing the accumulation of UV-damaged cell DNA and mutations (Saladi et al., 2010). Our results, in line with those of Fung et al. (2002), show different associations with risk for BCC by type of beverages. We have found no convincing documentation those beverage constituents other than alcohol affect the risk for NMSC. Furthermore, according to the World Cancer Research Fund (2007), the carcinogenetic factor in beverages is clearly alcohol itself, and there is no evidence that different types of beverages have markedly different effects on cancer risk. Hence, the different associations between types of beverage and BCC may be due to characteristics of the users or unmeasured confounding. High socioeconomic status, measured as education and disposable income, is associated with a high risk for BCC (Steding-Jessen et al., 2010), and socioeconomic status is known to be a strong predictor of beverage preference (Tjonneland et al., 2003). Hence, although we adjusted our analyses for duration of education, we cannot rule out the possibility that the observed increased risk for BCC associated with wine and spirits but not with beer is due to residual confounding by socioeconomic status. Furthermore, UV-related risk behavior may be associated with drinking, as a study has shown that the number of sunburns is associated with increasing alcohol intake (Mukamal, 2006). Interestingly, our study showed that the increased risk for BCC was not associated with the highest category of alcohol intake (X50 g per day), in particular for beer drinkers. Similar results for heavy drinkers have been reported previously (Corona et al., 2001; Fung et al., 2002; Freedman et al., 2003). The cause of the decreased risk among heavy drinkers has not been investigated directly, but both Fung et al. (2002) and Freedman et al. (2003) suggested that any effect of alcohol is overruled by other factors in heavy drinkers, such as a less sun-seeking behavior. Our results show no convincing associations between total alcohol intake and risk for SCC. The risk has been addressed 2724 Journal of Investigative Dermatology (2012), Volume 132

in only one cohort study (Ansems et al., 2008). This showed no overall increased risk for SCC, but subgroup analyses indicated that people with a previous skin cancer and a high intake of fortified wine were at increased risk. Both the results of Ansems et al. (2008) and ours are, however, limited by relatively small numbers of SCC cases (n ¼ 127 and n ¼ 198, respectively), reducing the possibility of detecting modest associations between alcohol intake and risk for SCC. More long-term follow-up studies are needed to clarify any association. Our study has several strengths. Data on all lifestyle risk factors were collected prospectively, minimizing the possibility of differential recall bias. Furthermore, loss to follow-up was virtually absent as a result of the precise linkage between the prospective Diet, Cancer and Health cohort and the Danish population-based registries. We were able to adjust for a number of potential confounders, including BMI, length of schooling, and use of HRT. Our study also had some potential limitations. The most important was that we were unable to adjust for UV dose acquired by each individual. We were, however, able to adjust for certain skin constitutional factors related to NMSC risk, such as skin reaction when exposed to strong sunlight, degree of freckling, and number of nevi. It is well established that people with skin that sunburns easily and tans poorly and people with a high degree of freckling and many nevi have a higher risk for NMSC (Lock-Andersen et al., 1999; Armstrong and Kricker, 2001). A further limitation of our study is that alcohol intake was captured only at baseline, and behavior might have changed between baseline and the end of follow-up; however, any misclassification of alcohol intake would not have been severe because it is unlikely that many people changed their alcohol intake dramatically during the followup period. Another potential limitation of our study is the reliability of the information about alcohol intake. Selfreported alcohol intake is generally considered reliable, although people may not know the correct portion sizes (Sommers, 2005). Any misclassification of alcohol intake in this study would most likely have been non-differential and would thus have resulted in underestimates of the strength of any associations, i.e., would have biased the estimated risks for BCC and SCC toward the null. In conclusion, the results of this large nationwide cohort study show an increased risk for BCC associated with current intake of wine and spirits and a non-monotonic association between intake of beer and BCC risk. For SCC, no convincing association with total alcohol intake was observed. Our results were limited by the fact that we were unable to adjust for the direct UV exposure of each person, and the risk estimates for SCC were imprecise because of the small number of cases. This limited our ability to detect even a modest association between intake of alcohol and risk for SCC, and additional long-term follow-up studies should be performed to confirm or refute our findings. In conjunction with a better biological understanding of alcohol as a potential risk factor for NMSC, these results will contribute to our understanding of the complex etiology of these cancers.

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

MATERIALS AND METHODS Study population The cohort has been described previously in detail (Tjonneland et al., 2007). Briefly, between 1993 and 1997, 80,996 men and 79,729 women living in the greater Copenhagen and Aarhus areas were invited to participate in the prospective Diet, Cancer and Health cohort. All the people invited were born in Denmark, aged 50–64 years and had no previous diagnosis of cancer. Each person was identified by his or her unique Danish personal identification number. A total of 57,054 persons corresponding to 35% of those invited were enrolled. The study was approved by the regional ethics committees on human studies in Copenhagen and Aarhus and by the Danish Data Protection Agency. At baseline, all participants received a detailed 192-item food frequency questionnaire by mail. Subsequently, the participants visited one of two study clinics in Copenhagen or Aarhus, where they completed a detailed lifestyle questionnaire with questions about alcohol intake and other lifestyle factors. Details of the development and validation of the questionnaires have been published previously (Overvad et al., 1991; Tjonneland et al., 1991). In the food frequency questionnaire, alcohol intake during the preceding year was recorded as the average frequency of intake of three types of beer in bottles (330 ml), wine in glasses (125 ml), fortified wine in drinks (60 ml), and spirits in drinks (30 ml). Red and white wine were not differentiated. The predefined responses were in 12 categories, ranging from ‘‘never’’ to ‘‘8 or more times a day’’. Alcohol content was calculated as follows: one bottle of light beer, 8.9 g ethanol; one bottle of regular beer, 12.2 g ethanol; one bottle of strong beer, 17.5 g ethanol; one glass of wine, 12.2 g ethanol; one drink of fortified wine, 9.3 g ethanol; and one drink of spirits, 9.9 g ethanol (Tjonneland et al., 2003). We obtained additional information on drinking patterns from the lifestyle questionnaire. The study subjects were asked about the frequency of alcohol drinking in the categories never, o1 per month, 1–3 times per month, once a week, 2–4 times per week, 5–6 times per week, and daily. From the lifestyle questionnaire, we also obtained information about potential confounders, such as skin reaction when exposed to strong sunlight, degree of freckling, number of nevi, length of education, parity, use of HRT, and smoking status. Weight and height were obtained at the two study clinics, and BMI was calculated as weight (kg) divided by height squared (m2). Information on vital status and migration was obtained by linking cohort members to the Central Population Registry by their personal identification numbers. To identify all incident NMSC cases diagnosed during the study period (1993–2007), each cohort member was linked to an already established NMSC database (Birch-Johansen et al., 2010), which lists incident cases of BCC and SCC diagnosed in Denmark in the period 1978–2007 and was formed by including the first recorded incident NMSC case from either the Danish Cancer Registry or the Danish Registry of Pathology. Each cohort member was followed from date of entry (first study center visit) until diagnosis of BCC or SCC or other malignancy, date of death, date of emigration or end of follow-up on 31 December 2007, whichever occurred first. A total of 571 cohort members (337 women and 234 men) who had any malignancy (including NMSC) reported to the Danish

Cancer Registry before study entry were excluded from data analysis. A further eight women and six men were excluded because they had answered only a small part of the lifestyle questionnaire. Furthermore, we excluded 490 women (1.6%) and 404 men (1.5%) for whom information on drinking patterns was missing and 685 women (2.3%) and 124 men (0.5%) for whom information on one or more potential confounders was missing, leaving 28,355 women (94.9%) and 26,411 men (97.2%) for analysis. The excluded persons were not systematically different from the included persons with regard to age at entry, median follow-up time, alcohol consumption, and potential confounders.

Statistical analysis We analyzed associations between the different alcohol exposure variables and combined as well as gender-specific BCC and SCC rates by use of the Cox proportional hazards model. In the Cox model, age was used as the underlying time axis to ensure that all analyses were based on comparisons of individuals of the same age. Furthermore, time under study was included as the time-dependent variable and modeled by a linear spline (Greenland, 1995) with boundaries at 1, 2, and 3 years after entry into the study cohort. All combined analysis models were stratified by gender to allow for separate underlying intensities in men and women. In the final analysis models, we included as potential confounders skin reaction when exposed to strong sunlight (categorical variable: redness, pain and blistering; redness, pain and peeling; redness then tan; only tan), degree of freckling (categorical variable: none, few, moderate, or many), number of nevi (categorical variable: none, few, moderate, or many), BMI (continuous variable), and duration of education (categorical variable: short (p7 years), medium (8–10 years), or long (410 years)). For models involving women, we also adjusted for menopausal status (categorical variable: pre- or post-menopausal) and use of HRT at baseline (categorical variable: never, past, or current use). These potential confounders were identified by automatic backward selection in a multivariate model for BCC risk; the model did not include use of alcohol at baseline. Potential confounders were retained in the final analysis model if their risk estimates were significant when mutually adjusted and/or if they influenced the other estimates. Other potential confounders that were considered but not included in the final models included pregnancy (ever/never), smoking status (categorical variable: never, past, or current smoking), and duration of HRT use. We tested all quantitative variables for linearity by using linear splines with knots placed at the quartiles (Greenland, 1995). No significant deviations from linearity were found for any of the variables. Two-sided 95% CIs for the estimated HRs were calculated with the Wald test of the Cox regression parameter (i.e., on the log (rate ratio) scale). We used the PHREG procedure, part of the SAS software package version 9.1 (SAS Institute, Cary, NC), for all statistical analyses. CONFLICT OF INTEREST The authors state no conflict of interest.

ACKNOWLEDGMENTS This research was supported by the Danish Cancer Society. The funding sources were not involved in the study design or data collection, analyses, interpretation of the results, the decision to submit the manuscript for publication, or writing the manuscript. The present work was performed in Copenhagen, Denmark.

www.jidonline.org 2725

A Jensen et al. Alcohol and Risk for Non-Melanoma Skin Cancer

REFERENCES Ansems TM, van der Pols JC, Hughes MC et al. (2008) Alcohol intake and risk of skin cancer: a prospective study. Eur J Clin Nutr 62:162–70 Armstrong BK, Kricker A (2001) The epidemiology of UV induced skin cancer. J Photochem Photobiol B 63:8–18 Birch-Johansen F, Jensen A, Mortensen L et al. (2010) Trends in the incidence of nonmelanoma skin cancer in Denmark 1978–2007: rapid incidence increase among young Danish women. Int J Cancer 127:2190–8 Corona R, Dogliotti E, D’Errico M et al. (2001) Risk factors for basal cell carcinoma in a Mediterranean population: role of recreational sun exposure early in life. Arch Dermatol 137:1162–8 Coutelle C, Hohn B, Benesova M et al. (2004) Risk factors in alcohol associated breast cancer: alcohol dehydrogenase polymorphism and estrogens. Int J Oncol 25:1127–32 Freedman DM, Sigurdson A, Doody MM et al. (2003) Risk of basal cell carcinoma in relation to alcohol intake and smoking. Cancer Epidemiol Biomarkers Prev 12:1540–3 Fung TT, Hunter DJ, Spiegelman D et al. (2002) Intake of alcohol and alcoholic beverages and the risk of basal cell carcinoma of the skin. Cancer Epidemiol Biomarkers Prev 11:1119–22 Green A, van der Pols J, Hunter D (2008) Skin cancer. In: Adami H, Hunter D, Trichopoulos D (eds) Textbook of Cancer Epidemiology, 2 edn. Oxford University Press: New York, 378–402

Milan T, Verkasalo PK, Kaprio J et al. (2003) Lifestyle differences in twin pairs discordant for basal cell carcinoma of the skin. Br J Dermatol 149:115–23 Mukamal KJ (2006) Alcohol consumption and self-reported sunburn: a crosssectional, population-based survey. J Am Acad Dermatol 55:584–9 Overvad K, Tjonneland A, Haraldsdottir J et al. (1991) Development of a semiquantitative food frequency questionnaire to assess food, energy and nutrient intake in Denmark. Int J Epidemiol 20:900–5 Poschl G, Seitz HK (2004) Alcohol and cancer. Alcohol 39:155–65 Sahl WJ, Glore S, Garrison P et al. (1995) Basal cell carcinoma and lifestyle characteristics. Int J Dermatol 34:398–402 Saladi RN, Nektalova T, Fox JL (2010) Induction of skin carcinogenicity by alcohol and ultraviolet light. Clin Exp Dermatol 35:7–11 Sommers MS (2005) Measurement of alcohol consumption: issues and challenges. Annu Rev Nurs Res 23:27–64 Steding-Jessen M, Birch-Johansen F, Jensen A et al. (2010) Socioeconomic status and non-melanoma skin cancer: a nationwide cohort study of incidence and survival in Denmark. Cancer Epidemiol 34:689–95 Tjonneland A, Olsen A, Boll K et al. (2007) Study design, exposure variables, and socioeconomic determinants of participation in Diet, Cancer and Health: a population-based prospective cohort study of 57,053 men and women in Denmark. Scand J Public Health 35:432–41

Greenland S (1995) Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 6:356–65

Tjonneland A, Overvad K, Haraldsdottir J et al. (1991) Validation of a semiquantitative food frequency questionnaire developed in Denmark. Int J Epidemiol 20:906–12

Kune GA, Bannerman S, Field B et al. (1992) Diet, alcohol, smoking, serum beta-carotene, and vitamin A in male nonmelanocytic skin cancer patients and controls. Nutr Cancer 18:237–44

Tjonneland A, Thomsen BL, Stripp C et al. (2003) Alcohol intake, drinking patterns and risk of postmenopausal breast cancer in Denmark: a prospective cohort study. Cancer Causes Control 14:277–84

Lock-Andersen J, Drzewiecki KT, Wulf HC (1999) Naevi as a risk factor for basal cell carcinoma in Caucasians: a Danish case-control study. Acta Derm Venereol 79:314–9

World Cancer Research Fund/American Institute for Cancer Research (2007) Food, Nutrition, Physical Activity, and the Prevention of Cancer: A Global Perspective. Washington DC

Madan V, Lear JT, Szeimies RM (2010) Non-melanoma skin cancer. Lancet 375:673–85

World Health Organization (2007) Global Status Report on Alcohol. World Health: Geneva

Merimsky O, Inbar M (1999) Alcohol intake-associated skin and mucosal cancer. Clin Dermatol 17:447–55

Young C (2009) Solar ultraviolet radiation and skin cancer. Occup Med (Lond) 59:82–8

2726 Journal of Investigative Dermatology (2012), Volume 132

Copyright of Journal of Investigative Dermatology is the property of Nature Publishing Group and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.