Abstract
Background: The association between history of keratinocyte carcinoma (KC, also known as nonmelanoma skin cancer) and risk of developing invasive melanoma has not been assessed comprehensively using prospective data.
Methods: We followed 91 846 women in the Nurses’ Health Study (NHS; 1984–2010), 114 918 women in the NHSII (1989–2011), and 48 946 men in the Health Professionals Follow-up Study (1986–2010) for diagnoses of KC and melanoma biennially. Cox proportional hazards models were used to compute hazard ratios (HRs) and 95% confidence intervals (CIs) of melanoma associated with history of KC, basal cell carcinoma (BCC), and squamous cell carcinoma (SCC). All statistical tests were two-sided.
Results: We documented 1949 melanomas, 38 842 BCCs, and 7462 SCCs over 6.4 million person-years of follow-up. After adjustment for other risk factors, a personal history of KC was associated with an increased risk of melanoma (meta-analysis HR = 2.22, 95% CI = 1.73 to 2.85). The association was more apparent among participants with a history of both BCC and SCC (HR = 3.40, 95% CI = 1.60 to 7.19) than among participants with a history of BCC only (HR = 2.20, 95% CI = 1.80 to 2.70) or SCC only (HR = 1.56, 95% CI = 0.98 to 2.46), and there was a strong risk-increasing trend associated with a higher number of reported KCs removed by surgery (Ptrend < .001). In women, KC history was more strongly associated with head/neck melanomas (HR = 4.17, 95% CI = 2.77 to 6.27) than with trunk or limb melanomas (both HRs < 2.50, Pheterogeneity = .04).
Conclusions: Our results provide novel insights for the relationship between KC history and risk of developing melanoma, which may be important for melanoma prevention.
Melanoma is the most lethal form of skin cancer, and its incidence has been increasing in the western population over the past several decades (1–3). Keratinocyte carcinomas (KCs), including cutaneous basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), are the most commonly diagnosed malignancies in the United States, with 3.5 million cases diagnosed among 2.2 million Americans each year (4). Both melanoma and KCs can be disfiguring, negatively affect quality of life, and cause substantial economic burden (5). Early case-control or registry-based studies have documented a positive association between KC history and melanoma (6–8), whereas prospective evidence for this relationship has been limited by either a small sample size or the inability to examine it in both women and men (9,10). Few studies have taken into account the subtypes of both KC (ie, BCC and SCC) and melanoma (eg, site-specific melanomas) simultaneously. Although BCC and SCC share similar environmental and host risk factors (eg, sun exposure and high sunburn susceptibility), these two types of skin cancer have heterogeneous disease characteristics. Furthermore, existing evidence suggests strong heterogeneity in risk profiles for melanomas occurring at different body sites (11–13), and it is unclear whether these melanomas are equally associated with history of KC. A substantial portion of KC patients are diagnosed with multiple KCs (4), whereas the impact of multiple prior-diagnosed KCs on subsequent risk of melanoma has been unknown. Epidemiologic studies in recent years have reported that associations between skin cancer risk and a number of other environmental factors besides sun exposure either may have protective (eg, caffeine) or potentiate (eg, citrus products) effects on the incidence of skin cancer (14–17). However, little is known about whether these potential skin cancer risk factors may modify the relationship between KC history and melanoma. Therefore, a prospective study with detailed risk factor information collected for KC and melanoma and sufficiently long follow-up is needed to more accurately and comprehensively investigate the relationship between KC history and melanoma.
Our previous study has reported an increased risk of cancer including melanoma in association with history of KC in the Nurses’ Health Study (NHS; 1984–2008) and Health Professionals Follow-up Study (HPFS; 1986–2008) (18). However, this previous report did not examine the detailed subgroups of KC in association with the risk of developing melanoma and melanoma subtypes, did not investigate the impact of multiple KCs on melanoma risk, and did not investigate the potential modification of the KC-melanoma association by other melanoma risk factors. Therefore, we conducted a further detailed study using data from the NHS (1984–2010) and HPFS (1986–2010) with extended follow-up and an additional cohort of women, the Nurses’ Health Study II (1989–2011).
Methods
Study Populations
The NHS was established in 1976 when 121 700 married, registered, female nurses age 30 to 55 years who resided in the United States at the time of enrollment responded to a baseline questionnaire that included questions about their medical history and lifestyle risk factors. The NHSII was established in 1989 when 116 430 registered female nurses age 25 to 42 years were enrolled using a mailed baseline questionnaire that inquired about medical history and lifestyle practices. The HPFS was established in 1986 when 51 529 US male health professionals age 40 to 75 years completed a baseline questionnaire on lifestyle, diet, and newly diagnosed diseases. Biennial questionnaires were used to collect data on environmental and lifestyle factors and disease outcomes in all three cohorts. Participants resided in around 15 states at enrollment and spread to all states in the United States during the follow-up. The institutional review boards of Partners Health Care System and Harvard School of Public Health approved this study. The participants’ completion and return of the self-administered questionnaires implied informed consent.
We inquired about history of KC for the first time in 1984 in the NHS, 1989 in the NHSII, and 1986 in the HPFS. Participants who did not return the baseline questionnaires or had a baseline history of melanoma or other nonskin cancers were excluded. Ninety-one thousand, eight hundred forty-six women in the NHS, 114 918 women in the NHSII, and 48 946 men in the HPFS remained in the study.
Assessment of History of Keratinocyte Carcinoma
Participants reported new diagnoses of SCC and BCC biennially starting in 1984 in the NHS, 1989 in the NHSII, and 1986 in the HPFS. Medical and pathological reports were obtained from participants who reported SCC to confirm the diagnosis. Medical records were not obtained for BCC cases. However, previous validation studies reported a high accuracy of self-reported BCC among subsets of cohort participants (96% in women and 84% in men) based on histopathological findings or medical records (19,20). Participants with confirmed SCC or self-reported BCC diagnosed before a given follow-up cycle were defined as having a history of KC at the beginning of that cycle. Participants were asked “How many squamous or basal cell carcinoma lesions have you ever had removed by surgery, cryotherapy or other means? (Exclude melanoma and benign lesions like moles or actinic keratoses)” in 2004 in NHS, 2005 in NHSII, and 2008 in HPFS with the following response categories: never, 1, 2–4, 5–10, or ≥11. The highest two categories were combined to maintain statistical power in data analysis.
Assessment of Melanoma Cases
Participants reported new diagnoses of melanoma biennially during the follow-up. Medical and pathological reports were collected from participants who reported melanoma and were reviewed by study physicians to confirm the diagnosis and retrieve information on tumor histology, including tumor location and Breslow thickness. Our primary end point was invasive melanoma, which was further classified into three site-specific categories according to tumor location: head/neck, trunk, and limbs.
Assessment of Covariates
Information on anthropometric and lifestyle factors for chronic diseases, including body height and weight, smoking status, and physical activity was asked and updated in the biennial follow-up questionnaires. Information on dietary factors, including consumption of citrus products, alcohol, and coffee, was repeatedly collected using a validated food frequency questionnaire at least every four years during the follow-up (17). Data on the following phenotypic and sun exposure–related factors were also collected in the follow-up questionnaires: ethnicities; family history of melanoma in first-degree relatives (parents, siblings, or offspring); natural hair color at an early age (age 21 years in NHS, and age 18 years in NHSII/HPFS); number of moles on extremity (arms in NHS/HPFS and legs in NHSII); skin reaction after prolonged sun exposure as a child/adolescent; number of severe or blistering sunburns; average time spent in direct sunlight in summer months since high school; regular use of sunscreen in summer months; and cumulative UV flux since baseline (a measure of long-term UV exposure) (14,17,21).
Statistical Analysis
In the analyses for KC history and its subtypes, participants contributed to follow-up time from the return month of the baseline questionnaire (June 1984 for NHS, June 1989 for NHSII, January 1986 for HPFS) to the month of the first diagnosis of melanoma or other nonskin cancers, month of death, or the end of follow-up (June 2010 for NHS, June 2011 for NHSII, January 2010 for HPFS), whichever came first. In the analysis for self-reported number of KCs, participants contributed to follow-up time from the return month of the questionnaires asking for the number of KCs removed by surgery (June 2004 for NHS, June 2005 for NHSII, January 2008 for HPFS).
We used Cox proportional hazards models stratified by age and follow-up intervals to estimate the age-adjusted and multivariable-adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) of overall melanoma and site-specific melanomas associated with KC history, which was treated as a time-dependent variable in the models. We verified the assumption of proportionality by modeling the interaction of KC history with follow-up time and found that inclusion of the time-dependent interaction did not statistically significantly improve the model fit for invasive melanoma (P = .25, likelihood ratio test, in the pooled test). KC history was further divided into the following three categories: BCC only, SCC only, and both BCC and SCC. Multivariable analyses were conducted with adjustment for age, follow-up interval, host risk score (quintiles), average time spent in direct sunlight since high school (<2, 2–5, or ≥ 6 hours/week), cumulative UV flux since baseline (quintiles), regular sunscreen use in summer months (yes or no), body mass index (<18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, or ≥ 35.0 kg/m2), smoking status (never, past, current with 1–14, 15–24, or 25 cigarettes/day), physical activity (quintiles), alcohol intake (0, 0.1–9.9, 10–19.9, 20.0–29.9, or ≥ 30.0 g/day), citrus consumption (quintiles), and caffeine intake (quintiles). To avoid overadjustment, a host risk score was created for each participant using cohort-derived hazard ratios associated with each of the six host risk factors of melanoma (ie, ethnicities, family history of melanoma, natural hair color, number of moles on arms, skin reaction after prolonged sun exposure as a child/adolescent, and number of severe or blistering sunburns) (21) and was adjusted in the models as quintiles. A meta-analysis approach was used to combine the risk estimates in different cohorts using a random effect model. Subgroup analyses were performed for several adjusted variables in the presence of appreciable risk difference over subgroups. Heterogeneity for associations between subgroups of KC and risk of melanoma or melanoma subtypes and potential interaction between history of KC and other skin cancer risk factors were examined with the use of the Q statistic (22). We used the t test to compare the difference in Breslow thickness between melanoma cases with and without history of KC. We used SAS software version 9.2 (SAS Institute Inc., Cary, NC) for all statistical analyses. All statistical tests were two-sided, and the statistical significance level was set at a P value of .05.
Results
We documented 1949 invasive melanomas in three cohorts over 6.4 million person-years of follow-up (772 cases/2 795 275 person-years in NHS, 552 cases/2 573 152 person-years in NHSII, 625 cases/1 009 275 person-years in HPFS). In women, 130 melanomas (9.8%) occurred on head/neck, 421 (31.8%) on trunk, and 739 (55.8%) on limbs; in men, 158 melanomas (25.2%) occurred on head/neck, 268 (42.8%) on trunk, and 129 (20.6%) on limbs. During the follow-up, 38 842 participants (28 074 women and 10 768 men) developed BCC, and 7462 participants (5038 women and 2424 men) developed SCC. Participants with a history of KC appeared to be older and were more likely to have host risk factors (eg, family history of melanoma, natural red/blonde hair, and presence of extremity moles) and higher exposure to UV flux and longer average time spent in direct sunlight (Table 1).
Table 1.
Characteristics of person-years according to history of keratinocyte carcinoma in the Nurses’ Health Study (1984–2010), Nurses’ Health Study II (1989–2011), and Health Professionals Follow-up Study (1986–2010)*
| Characteristic | NHS |
NHS II |
HPFS |
|||
|---|---|---|---|---|---|---|
| No (n = 1 065 559) | Yes (n = 60 642) | No (n = 1 286 044) | Yes (n = 38 040) | No (n = 474 681) | Yes (n = 47 154) | |
| Age, mean (SD), y | 53.1 (10.0) | 61.2 (8.8) | 42.9 (8.0) | 48.9 (7.1) | 60.6 (10.8) | 69.0 (9.7) |
| Body mass index, Mean (SD), kg/m2 | 26.3 (5.3) | 25.5 (5.0) | 26.0 (6.1) | 25.1 (5.4) | 25.8 (4.3) | 25.8 (3.9) |
| Physical activity level, Mean (SD), metabolic-equivalent h/wk | 17.2 (22.4) | 19.3 (24.7) | 21.7 (29.4) | 25.1 (35.9) | 33.1 (39.6) | 40.5 (41.0) |
| Current smoking, % | 13.9 | 12.6 | 10.0 | 9.4 | 6.4 | 4.6 |
| Past smoking, % | 41.2 | 45.2 | 24.4 | 27.4 | 39.1 | 38.1 |
| Caucasian ancestry, % | 97.2 | 99.6 | 95.4 | 99.1 | 95.2 | 97.8 |
| Family history of melanoma, % | 9.5 | 12.3 | 11.6 | 16.5 | 3.9 | 5.7 |
| Red/blonde hair, % | 15.0 | 21.8 | 19.4 | 28.5 | 13.0 | 16.6 |
| Arm with moles, % | 35.7 | 44.1 | 50.1 | 56.2 | 31.3 | 35.7 |
| Painful burn/blisters reaction as a child/adolescent, % | 14.0 | 23.3 | 23.2 | 35.1 | 23.5 | 34.2 |
| Blistering sunburns, Mean (SD) | 8.4 (6.9) | 11.3 (7.6) | 1.9 (2.3) | 2.9 (2.9) | 12.7 (12.1) | 15.7 (12.9) |
| Annual UV flux (x10−4 RB count), Mean (SD) | 123.5 (25.6) | 127.8 (27.4) | 126.0 (25.1) | 130.7 (26.5) | 130.6 (27.6) | 136.5 (28.8) |
| Average time spent in direct sunlight in summer months since high school, Mean (SD), h/wk | 5.4 (2.9) | 5.7 (3.0) | 4.4 (1.6) | 4.5 (1.5) | 10.2 (6.0) | 10.6 (6.1) |
| Regular use of sunscreen in summer months, % | 24.0 | 35.3 | 81.1 | 88.4 | 58.6 | 74.7 |
| Citrus consumption, Mean (SD), serving/d | 0.8 (0.6) | 0.8 (0.6) | 0.5 (0.6) | 0.6 (0.5) | 0.9 (0.8) | 0.9 (0.7) |
| Caffeine intake, Mean (SD), mg/d | 276.1 (198.1) | 265.7 (191.2) | 229.4 (190.9) | 217.0 (169.9) | 221.3 (203.7) | 207.3 (184.4) |
| Alcohol intake, Mean (SD), g/d | 5.9 (9.7) | 7.1 (10.2) | 3.5 (6.1) | 4.1 (6.1) | 11.1 (14.5) | 11.6 (13.7) |
All values except age have been standardized to the age distribution of the study population. HPFS = Health Professionals Follow-up Study; NHS = Nurses’ Health Study; RB = Robertson-Berger; UV = ultraviolet.
After adjustment for other skin cancer risk factors, a personal history of KC was strongly associated with increased melanoma risk in all three cohorts (Table 2, meta-analysis multivariable-adjusted HR = 2.22, 95% CI = 1.73 to 2.85). Analyses stratified by history of BCC and/or SCC suggested a substantially increased risk of melanoma among those who had been diagnosed with both BCC and SCC (meta-analysis HR = 3.40, 95% CI = 1.60 to 7.19), followed by the risk among those with history of BCC only (meta-analysis HR = 2.20, 95% CI = 1.80 to 2.70) and SCC only (meta-analysis HR = 1.56, 95% CI = 0.98 to 2.46) (Pheterogeneity = .24) (Table 2). Melanoma cases with a personal history of KC appeared to have a thicker tumor compared with cases without the history in the NHS and HPFS, but not in the NHSII (Supplementary Table 1, available online).
Table 2.
Risk of melanoma according to history of keratinocyte carcinoma in the Nurses’ Health Study (1984–2010), Nurses’ Health Study II (1989–2011), and Health Professionals Follow-up Study (1986–2010)
| Cohort/History of KC | No. of cases | No. of person-years | Age-adjusted HR (95% CI) | Multivariable-adjusted HR (95% CI)* | Pheterogeneity† |
|---|---|---|---|---|---|
| NHS (1984–2010) | |||||
| No | 581 | 2 606 050 | 1.00 | 1.00 | |
| Yes | 191 | 189 225 | 3.19 (2.69 to 3.78) | 2.68 (2.26 to 3.18) | |
| BCC only | 157 | 170 355 | 2.97 (2.48 to 3.56) | 2.52 (2.09 to 3.02) | |
| SCC only | 6 | 8386 | 2.06 (0.92 to 4.60) | 1.82 (0.81 to 4.08) | |
| BCC and SCC | 28 | 10 484 | 7.36 (5.00 to 10.8) | 5.61 (3.80 to 8.28) | .001 |
| NHS II (1989–2011) | |||||
| No | 507 | 2 500 468 | 1.00 | 1.00 | |
| Yes | 45 | 72 742 | 2.92 (2.14 to 3.98) | 2.15 (1.57 to 2.94) | |
| BCC only | 43 | 66 133 | 3.07 (2.24 to 4.22) | 2.28 (1.66 to 3.13) | |
| SCC only | 1 | 4634 | 0.98 (0.14 to 7.00) | 0.70 (0.10 to 5.02) | |
| BCC and SCC | 1 | 1974 | 2.40 (0.34 to 17.1) | 1.45 (0.20 to 10.4) | .47 |
| HPFS (1986–2010) | |||||
| No | 490 | 918 592 | 1.00 | 1.00 | |
| Yes | 135 | 90 684 | 2.26 (1.84 to 2.77) | 1.85 (1.51 to 2.28) | |
| BCC only | 109 | 75 618 | 2.23 (1.79 to 2.77) | 1.84 (1.48 to 2.30) | |
| SCC only | 12 | 9154 | 1.92 (1.07 to 3.43) | 1.54 (0.86 to 2.75) | |
| BCC and SCC | 14 | 5912 | 3.26 (1.88 to 5.64) | 2.43 (1.40 to 4.21) | .52 |
| Meta-analysis for women (NHS/NHS II) | |||||
| No | 1088 | 5 106 462 | 1.00 | 1.00 | |
| Yes | 236 | 261 965 | 3.13 (2.69 to 3.63) | 2.50 (2.04 to 3.06) | |
| BCC only | 200 | 236 488 | 2.99 (2.56 to 3.50) | 2.45 (2.09 to 2.88) | |
| SCC only | 7 | 13 019 | 1.85 (0.88 to 3.89) | 1.58 (0.75 to 3.34) | |
| BCC and SCC | 29 | 12 458 | 6.47 (3.21 to 13.0) | 4.08 (1.33 to 12.6) | .35 |
| Meta-analysis for women and men (NHS/NHS II/HPFS) | |||||
| No | 1578 | 6 025 054 | 1.00 | 1.00 | |
| Yes | 371 | 352 649 | 2.76 (2.19 to 3.48) | 2.22 (1.73 to 2.85) | |
| BCC only | 309 | 312 106 | 2.71 (2.20 to 3.32) | 2.20 (1.80 to 2.70) | |
| SCC only | 19 | 22 173 | 1.89 (1.20 to 2.99) | 1.56 (0.98 to 2.46) | |
| BCC and SCC | 43 | 18 370 | 4.65 (2.29 to 9.45) | 3.40 (1.60 to 7.19) | .24 |
Multivariable analyses were adjusted for age, follow-up interval, host risk score (quintiles), average time spent in direct sunlight since high school (<2, 2–5, or ≥ 6 hours/week), cumulative UV flux since baseline (quintiles), regular sunscreen use in summer months (yes or no), body mass index (<18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, or ≥ 35.0 kg/m2), smoking status (never, past, current with 1–14, 15–24, or 25 cigarettes/day), physical activity (quintiles), alcohol intake (0, 0.1–9.9, 10–19.9, 20.0–29.9, or ≥ 30.0 g/day), citrus consumption (quintiles), and caffeine intake (quintiles). Meta-analyses for different cohorts were performed using the random effect model. CI = confidence interval; HPFS = Health Professionals Follow-up Study; HR = hazard ratio; NHS = Nurses’ Health Study.
Two-sided Pheterogeneity values between history of BCC only, history of SCC only, and history of both BCC and SCC were calculated with the use of the Q statistic (22).
There were statistically significantly increasing trends in melanoma risk over the categories of self-reported KCs removed by surgery in all three cohorts, especially in the NHS and NHSII (Table 3). The meta-analysis multivariable-adjusted hazard ratios for melanoma were 1.98 (95% CI = 1.51 to 2.59) for participants with only one KC, 2.23 (95% CI = 1.58 to 3.14) for two to four KCs, and 3.62 (95% CI = 2.46 to 5.32) for five or more KCs (Ptrend < .001). Analyses for site-specific melanomas suggested that the association with KC history was slightly apparent for melanomas occurring on head/neck, followed by melanomas occurring on limbs and trunk, and the difference was mostly driven by women (Table 4). In women, the multivariable-adjusted hazard ratio for head/neck melanoma (4.17, 95% CI = 2.77 to 6.27) was stronger than that for trunk melanoma (HR = 2.22, 95% CI = 1.47 to 3.37, Pheterogeneity = .03) or limb melanoma (HR = 2.40, 95% CI = 1.96 to 2.94, Pheterogeneity = .02).
Table 3.
Risk of melanoma according to self-reported number of keratinocyte carcinomas ever removed by surgery in the Nurses’ Health Study (2004–2010), Nurses’ Health Study II (2005–2011), and Health Professionals Follow-up Study (2008–2010)
| Cohort/No. of KC | No. of cases | No. of person-years | Age-adjusted HR (95% CI) | Multivariable-adjusted HR (95% CI)* |
|---|---|---|---|---|
| NHS (2004–2010) | ||||
| No | 165 | 615 819 | 1.00 | 1.00 |
| 1 | 41 | 78 056 | 1.97 (1.39 to 2.77) | 1.78 (1.26 to 2.52) |
| 2–4 | 31 | 54 339 | 2.14 (1.45 to 3.15) | 1.79 (1.21 to 2.65) |
| ≥5 | 20 | 15 832 | 4.78 (2.99 to 7.63) | 3.65 (2.27 to 5.87) |
| Ptrend† | <.001 | <.001 | ||
| NHS II (2005–2011) | ||||
| No | 144 | 549 649 | 1.00 | 1.00 |
| 1 | 18 | 26 492 | 2.58 (1.58 to 4.21) | 2.17 (1.33 to 3.56) |
| 2–4 | 13 | 13 677 | 3.63 (2.05 to 6.41) | 2.83 (1.59 to 5.03) |
| ≥5 | 6 | 3528 | 6.32 (2.79 to 14.32) | 4.31 (1.89 to 9.85) |
| Ptrend† | <.001 | <.001 | ||
| HPFS (2008–2010) | ||||
| No | 12 | 24 279 | 1.00 | 1.00 |
| 1 | 8 | 4922 | 3.20 (1.30 to 7.89) | 3.01 (1.21 to 7.49) |
| 2–4 | 10 | 5007 | 3.57 (1.52 to 8.39) | 3.09 (1.29 to 7.39) |
| ≥5 | 5 | 2493 | 3.38 (1.17 to 9.77) | 2.53 (0.85 to 7.50) |
| Ptrend† | .002 | .01 | ||
| Meta-analysis for women (NHS/NHS II) | ||||
| No | 309 | 1 165 468 | 1.00 | 1.00 |
| 1 | 59 | 104 548 | 2.15 (1.62 to 2.85) | 1.90 (1.43 to 2.52) |
| 2–4 | 44 | 68 016 | 2.67 (1.60 to 4.44) | 2.14 (1.38 to 3.31) |
| ≥5 | 26 | 19 360 | 5.12 (3.41 to 7.68) | 3.81 (2.52 to 5.75) |
| Ptrend† | <.001 | <.001 | ||
| Meta-analysis for women and men (NHS/NHS II/HPFS) | ||||
| No | 321 | 1 189 747 | 1.00 | 1.00 |
| 1 | 67 | 109 470 | 2.23 (1.70 to 2.91) | 1.98 (1.51 to 2.59) |
| 2–4 | 54 | 73 023 | 2.76 (1.89 to 4.03) | 2.23 (1.58 to 3.14) |
| ≥5 | 31 | 21 853 | 4.85 (3.32 to 7.09) | 3.62 (2.46 to 5.32) |
| Ptrend† | <.001 | <.001 | ||
Multivariable analyses were adjusted for covariate age, follow-up interval, host risk score (quintiles), average time spent in direct sunlight since high school (<2, 2–5, or ≥ 6 hours/week), cumulative UV flux since baseline (quintiles), regular sunscreen use in summer months (yes or no), body mass index (<18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, or ≥ 35.0 kg/m2), smoking status (never, past, current with 1–14, 15–24 or 25 cigarettes/day), physical activity (quintiles), alcohol intake (0, 0.1–9.9, 10–19.9, 20.0–29.9, or ≥ 30.0 g/d), citrus consumption (quintiles), and caffeine intake (quintiles). Meta-analyses for different cohorts were performed using the random effect model. CI = confidence interval; HPFS = Health Professionals Follow-up Study; HR = hazard ratio; NHS = Nurses’ Health Study.
Two-sided Ptrend values were calculated with self-reported number of keratinocyte carcinomas ever removed by surgery as an ordinal variable in the model.
Table 4.
Risk of site-specific melanoma according to history of keratinocyte carcinoma in the Nurses’ Health Study (1984–2010), Nurses’ Health Study II (1989–2011), and Health Professionals Follow-up Study (1986–2010)
| Cohort/history of KC | No. of person-years | Head/neck |
Trunk |
Limbs |
Pheterogeneity† | |||
|---|---|---|---|---|---|---|---|---|
| No. of cases | Multivariable- adjusted HR (95% CI)* | No. of cases | Multivariable- adjusted HR (95% CI)* | No. of cases | Multivariable- adjusted HR (95% CI)* | |||
| NHS (1984–2010) | ||||||||
| No | 2 606 050 | 48 | 1.00 | 176 | 1.00 | 341 | 1.00 | |
| Yes | 189 225 | 32 | 4.45 (2.78 to 7.13) | 52 | 2.59 (1.87 to 3.57) | 102 | 2.44 (1.94 to 3.08) | .07 |
| NHS II (1989–2011) | ||||||||
| No | 2 500 412 | 43 | 1.00 | 181 | 1.00 | 270 | 1.00 | |
| Yes | 72 740 | 7 | 3.40 (1.50 to 7.73) | 12 | 1.65 (0.91 to 3.00) | 26 | 2.28 (1.51 to 3.45) | .37 |
| HPFS (1986–2010) | ||||||||
| No | 918 592 | 114 | 1.00 | 216 | 1.00 | 102 | 1.00 | |
| Yes | 90 684 | 44 | 1.95 (1.34 to 2.83) | 52 | 1.95 (1.40 to 2.71) | 27 | 1.58 (1.00 to 2.51) | .73 |
| Meta-analysis for women (NHS/NHS II) | ||||||||
| No | 5 106 555 | 91 | 1.00 | 357 | 1.00 | 611 | 1.00 | |
| Yes | 261 953 | 39 | 4.17 (2.77 to 6.27) | 64 | 2.22 (1.47 to 3.37) | 128 | 2.40 (1.96 to 2.94) | .04 |
| Meta-analysis for women and men (NHS/NHS II/HPFS) | ||||||||
| No | 6 027 262 | 205 | 1.00 | 573 | 1.00 | 713 | 1.00 | |
| Yes | 352 919 | 83 | 3.01 (1.67 to 5.43) | 116 | 2.16 (1.71 to 2.73) | 155 | 2.20 (1.75 to 2.76) | .59 |
Multivariable analyses were adjusted for covariate age, follow-up interval, host risk score (quintiles), average time spent in direct sunlight since high school (<2, 2–5, or ≥ 6 hours/week), cumulative UV flux since baseline (quintiles), regular sunscreen use in summer months (yes or no), body mass index (<18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, or ≥ 35.0 kg/m2), smoking status (never, past, current with 1–14, 15–24 or 25 cigarettes/day), physical activity (quintiles), alcohol intake (0, 0.1–9.9, 10–19.9, 20.0–29.9, or ≥ 30.0 g/d), citrus consumption (quintiles), and caffeine intake (quintiles). Meta-analyses for different cohorts were performed using the random effect model. CI = confidence interval; HPFS = Health Professionals Follow-up Study; HR = hazard ratio; NHS = Nurses’ Health Study.
Two-sided Pheterogeneity values were calculated with the use of the Q statistic (22).
Subgroup analyses showed that the association between KC history and melanoma tended to be more apparent among participants who lived in areas with low annual UV flux at residence (Table 5). The meta-analysis multivariable-adjusted HR was slightly higher among participants with lower annual UV flux (2.61, 95% CI = 1.87 to 3.65) than among participants with higher annual UV flux at residence (1.83, 95% CI = 1.52 to 2.20, Pinteraction = .07). This association pattern was similar in women and men (data not shown).
Table 5.
Risk of melanoma according to history of keratinocyte carcinoma in subgroups in the Nurses’ Health Study (1984–2010), Nurses’ Health Study II (1989–2011), and Health Professionals Follow-up Study (1986–2010)
| Covariate/history of KC | No. of cases | No. of person-years | Multivariable- adjusted HR (95% CI)* | No. of cases | No. of person-years | Multivariable- adjusted HR (95% CI)* | Pinteraction† |
|---|---|---|---|---|---|---|---|
| Annual UV flux | |||||||
| Below median | Above median | ||||||
| No | 867 | 3 554 871 | 1.00 | 707 | 2 463 036 | 1.00 | |
| Yes | 209 | 180 875 | 2.61 (1.87 to 3.65) | 162 | 172 629 | 1.83 (1.52 to 2.20) | .07 |
| BCC only | 181 | 162 002 | 2.70 (2.12 to 3.44) | 128 | 150 690 | 1.74 (1.42 to 2.12) | .006 |
| SCC only | 7 | 8256 | 1.61 (0.76 to 3.40) | 12 | 11 854 | 1.68 (0.94 to 3.00) | .90 |
| BCC and SCC | 21 | 8448 | 3.19 (0.85 to 12.0) | 22 | 8984 | 3.49 (2.23 to 5.44) | .88 |
| Regular sunscreen use in summer months | |||||||
| No | Yes | ||||||
| No | 555 | 2 466 637 | 1.00 | 853 | 2 708 028 | 1.00 | |
| Yes | 148 | 148 930 | 2.45 (2.03 to 2.97) | 193 | 168 160 | 2.25 (1.50 to 3.38) | .72 |
| BCC only | 113 | 132 123 | 2.28 (1.63 to 3.17) | 167 | 147 206 | 2.28 (1.62 to 3.22) | .98 |
| SCC only | 7 | 8191 | 1.74 (0.82 to 3.72) | 12 | 11 721 | 1.79 (0.97 to 3.30) | .95 |
| BCC and SCC | 28 | 7653 | 6.97 (4.66 to 10.5) | 14 | 9234 | 2.30 (0.72 to 7.33) | .07 |
| Citrus consumption | |||||||
| Below median | Above median | ||||||
| No | 673 | 2 695 163 | 1.00 | 817 | 2 613 924 | 1.00 | |
| Yes | 188 | 171 027 | 2.61 (2.01 to 3.38) | 178 | 176 238 | 1.86 (1.28 to 2.70) | .15 |
| BCC only | 155 | 150 887 | 2.57 (1.92 to 3.43) | 149 | 155 325 | 1.87 (1.35 to 2.58) | .15 |
| SCC only | 12 | 11 047 | 2.31 (1.29 to 4.16) | 7 | 11 350 | 1.08 (0.51 to 2.28) | .12 |
| BCC and SCC | 21 | 9094 | 3.96 (2.00 to 7.83) | 22 | 9563 | 4.11 (2.64 to 6.39) | .93 |
| Caffeine intake | |||||||
| Below median | Above median | ||||||
| No | 787 | 2 632 766 | 1.00 | 688 | 2 591 513 | 1.00 | |
| Yes | 192 | 187 821 | 1.94 (1.38 to 2.73) | 172 | 155 755 | 2.56 (2.11 to 3.10) | .17 |
| BCC only | 164 | 165 103 | 1.99 (1.48 to 2.67) | 138 | 137 561 | 2.41 (1.99 to 2.93) | .30 |
| SCC only | 7 | 12 090 | 0.95 (0.45 to 2.02) | 12 | 7802 | 2.58 (1.44 to 4.64) | .04 |
| BCC and SCC | 21 | 9655 | 3.23 (2.00 to 5.23) | 22 | 7972 | 4.81 (2.32 to 9.97) | .37 |
Multivariable analyses were adjusted for covariate age, follow-up interval, host risk score (quintiles), average time spent in direct sunlight since high school (<2, 2–5, or ≥ 6 hours/week), cumulative UV flux since baseline (quintiles), regular sunscreen use in summer months (yes or no), body mass index (<18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, or ≥ 35.0 kg/m2), smoking status (never, past, current with 1–14, 15–24 or 25 cigarettes/day), physical activity (quintiles), alcohol intake (0, 0.1–9.9, 10–19.9, 20.0–29.9, or ≥ 30.0 g/d), citrus consumption (quintiles), and caffeine intake (quintiles). Meta-analyses for different cohorts were performed using the random effect model. BCC = basal cell carcinoma; CI = confidence interval; HR = hazard ratio; SCC = squamous cell carcinoma; UV = ultraviolet.
Two-sided Pinteraction values were calculated with the use of the Q statistic comparing the subgroup-specific multivariable-adjusted hazard ratios calculated by meta-analysis (22).
Discussion
In the present study, we provided a detailed evaluation for the association between personal history of KC and risk of invasive melanoma based on data from three large cohorts of women and men. After adjustment for other skin cancer risk factors, we found that participants with a history of both BCC and SCC have a more apparently increased risk of melanoma than participants with a history of BCC only or SCC only. There was a strong increasing trend in melanoma risk associated with a higher number of reported KCs removed by surgery. Analyses for site-specific melanomas suggest that head/neck melanomas were more strongly associated with KC history, particularly in women.
Our results on the increased risk of melanoma in association with personal history of KC were consistent with previous reports, except that the risk estimate tended to be lower in the present study (multivariable-adjusted HR = 2.22) than those reported in previous case-control or registry-based studies (HR/meta-analysis relative risk > 2.70) (6–8) and in a prospective study with a total of 19 174 individuals and a 16-year follow-up (multivariable-adjusted relative risk = 7.94) (9). The modest associations reported in our study may be partly explained by the adjustment for a number of potential risk factors. In the present study, we were able to investigate the association among both women and men with a larger sample size and further differentiate between BCC and SCC based on detailed cohort follow-up information. We found that participants with a history of both BCC and SCC had a more apparently increased risk of melanoma when compared with participants with a history of BCC or SCC alone, thus adding further evidence to our previous report on the association between KC history and melanoma in the NHS and HPFS (18). However, the statistical test for the heterogeneity between history of BCC or SCC alone and history of both BCC and SCC did not reach statistical significance, and further studies are needed to confirm our findings. We also found that participants who reported a higher number of KCs had a substantially increased risk of melanoma when compared with those who reported no KC removed by surgery. Nevertheless, it is possible that individuals with a prior history of KC or multiple KCs may be more aware of their health conditions and thus seek more frequent medical examinations to allow for early detection of melanoma, resulting in overdiagnosis of melanoma cases among individuals with a history of KC. Our data also partially support the possibility of detection bias, as shown by a slightly lower mean thickness of melanomas among participants with a history of KC in the NHS and HPFS (Table S1). However, melanomas among the NHSII participants with a history of KC had a higher mean thickness than melanomas among those without the history, suggesting potential heterogeneity in the detection bias between older participants (NHS and HPFS) and younger participants (NHSII), which merits further investigation. In addition, there is also a possibility that some of these melanomas would be indolent, or that they would regress naturally, and future studies are needed to clarify these issues.
A strong literature exists for the different risk factor profiles for melanomas occurring at different body sites. Head/neck melanomas have been more closely associated with chronic sun exposure, trunk melanomas have been more closely associated with intermittent sun exposure (eg, severe sunburn) and host risk factors (eg, presence of moles/nevi), and limb melanomas appear to have an intermediate risk factor profile (11–13). Our findings are consistent with the existing literature, as demonstrated by the more apparent association between history of KC and head/neck melanomas when compared with those between history of KC and trunk and limb melanomas. Chronic UV exposure is the major established environmental risk factor for KC, and thus a history of KC may imply a history of high continuous UV exposure. However, the difference for site-specific melanomas was apparent in women but not in men, which may be associated with the different sun exposure patterns between genders. In women, head/neck receives the highest UV radiation, followed by that received by limbs, and trunk receives the least exposure. This gradient in sun exposure over the body may be less apparent in men and may have diminished the risk difference for melanomas occurring at different body sites.
In the model analysis, we controlled for a set of host, environmental, lifestyle, and dietary factors, most of which have been associated with risk of melanoma or KC in previous epidemiologic studies (14–16,21,23–25). Interestingly, the association was more apparent among participants who lived in areas with lower annual UV flux. High levels of cumulative UV flux have been associated with increased risk of melanoma in previous studies (21,26). It is thus plausible that having a history of KC may represent a more susceptible state and could raise the risk of melanoma more substantially among individuals with lower UV exposure.
Our study has a number of strengths, including its prospective design, large sample size, long-term follow-up, inclusion of both female and male cohorts, a geographically diverse population from all over the United States, accurate case ascertainment on melanoma, and the ability to control for a number of potential skin cancer risk factors based on the detailed cohort follow-up. These strengths are important for a more accurate evaluation of the association between KC history and subsequent risk of melanoma. Our study also has several limitations. First, our study participants mostly comprised well-educated, white health professionals, which may not be a representative sample of the general population. However, restricting the sample to health professionals also reduces the possibility of introducing a confounding effect associated with socioeconomic status. Second, although BCC self-reports have a high accuracy in our study cohorts, we did not validate BCCs using medical records in the study participants because of the vast cases and high costs. Therefore we do not have specific data on body site distribution of BCC and could not evaluate the association between site-specific KC and melanoma risk.
In summary, our study confirmed that a personal history of KC was strongly associated with increased risk of invasive melanoma in both women and men based on data from three cohort studies. Our study provided several novel insights into the existing literature on KC history and melanoma risk, which may have useful health implications for melanoma prevention in clinical practices. For example, individuals who have been diagnosed with KC should be aware of the potentially increased risk of melanoma, and physicians need to pay more attention to these high-risk individuals, especially those who were previously diagnosed with both BCC and SCC or who had multiple KCs. In women, special attention should be paid to head/neck melanomas among those with a history of KC.
Funding
This work was supported by National Institutes of Health (grants UM1 CA186107, P01 CA87969, UM1 CA176726, UM1 CA167552, and R01 CA137365). Wen-Qing Li was supported by the Research Career Development Award of Dermatology Foundation and the Richard B. Salomon Faculty Research Award of Brown University.
Notes
The funding source had no role in the design of the study; the data collection, analysis, or interpretation; the manuscript writing; or the decision to submit the manuscript for publication.
We would like to thank the participants and staff of the cohort studies for their valuable contributions, as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. Dr. Wu and Dr. Qureshi had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Dr. Qureshi has received honoraria that have been donated to charity from Abbvie, Amgen, Centers for Disease Control, Janssen, Merck, Novartis, Pfizer (consultant), and Amgen (investigator). Dr. Qureshi is an investigator for Sanofi/Reneron with no financial compensation.
Supplementary Material
References
- 1. American Cancer Society. Cancer Facts & Figures 2014. Atlanta, GA: American Cancer Society; 2014. [Google Scholar]
- 2. Parkin DM, Bray F, Ferlay J, Pisani P.. Global cancer statistics, 2002. CA Cancer J Clin. 2005;55:74–108. [DOI] [PubMed] [Google Scholar]
- 3. Geller AC, Clapp RW, Sober AJ, et al. Melanoma epidemic: An analysis of six decades of data from the Connecticut Tumor Registry. J Clin Oncol. 2013;31:4172–4178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. American Cancer Society. Skin cancer: Basal and squamous cell. What are the key statistics about basal and squamous cell skin cancers? 2014. http://www.cancer.org/cancer/skincancer-basalandsquamouscell/detailedguide/skin-cancer-basal-and-squamous-cell-key-statistics. Accessed November 26, 2014.
- 5. Guy GP, Ekwueme DU.. Years of potential life lost and indirect costs of melanoma and non-melanoma skin cancer: A systematic review of the literature. PharmacoEconomics. 2011;29:863–874. [DOI] [PubMed] [Google Scholar]
- 6. Rees JR, Zens MS, Gui J, Celaya MO, Riddle BL, Karagas MR.. Non melanoma skin cancer and subsequent cancer risk. PloS One. 2014;9:e99674. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Nugent Z, Demers AA, Wiseman MC, Mihalcioiu C, Kliewer EV.. Risk of second primary cancer and death following a diagnosis of nonmelanoma skin cancer. Cancer Epidemiol Biomarkers Prev. 2005;14:2584–2590. [DOI] [PubMed] [Google Scholar]
- 8. Wheless L, Black J, Alberg AJ.. Nonmelanoma skin cancer and the risk of second primary cancers: A systematic review. Cancer Epidemiol Biomarkers Prev. 2010;19:1686–1695. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Chen J, Ruczinski I, Jorgensen TJ, et al. Nonmelanoma skin cancer and risk for subsequent malignancy. J Natl Cancer Inst. 2008;100:1215–1222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Rosenberg CA, Khandekar J, Greenland P, Rodabough RJ, McTiernan A.. Cutaneous melanoma in postmenopausal women after nonmelanoma skin carcinoma: The Women's Health Initiative Observational Study. Cancer. 2006;106:654–663. [DOI] [PubMed] [Google Scholar]
- 11. Whiteman DC, Stickley M, Watt P, Hughes MC, Davis MB, Green AC.. Anatomic site, sun exposure, and risk of cutaneous melanoma. J Clin Oncol. 2006;24:3172–3177. [DOI] [PubMed] [Google Scholar]
- 12. Olsen CM, Zens MS, Stukel TA, et al. Nevus density and melanoma risk in women: A pooled analysis to test the divergent pathway hypothesis. Int J Cancer. 2009;124:937–944. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Kvaskoff M, Pandeya N, Green AC, et al. Site-specific determinants of cutaneous melanoma: A case-case comparison of patients with tumors arising on the head or trunk. Cancer Epidemiol Biomarkers Prev. 2013;22:2222–2231. [DOI] [PubMed] [Google Scholar]
- 14. Wu S, Han J, Song F, et al. Caffeine intake, coffee consumption, and risk of cutaneous malignant melanoma. Epidemiology. 2015;26:898–908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Loftfield E, Freedman ND, Graubard BI, et al. Coffee drinking and cutaneous melanoma risk in the NIH-AARP diet and health study. J Natl Cancer Inst. 2015;107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Rota M, Pasquali E, Bellocco R, et al. Alcohol drinking and cutaneous melanoma risk: A systematic review and dose-risk meta-analysis. Br J Dermatol. 2014;170:1021–1028. [DOI] [PubMed] [Google Scholar]
- 17. Wu S, Han J, Feskanich D, et al. Citrus consumption and risk of cutaneous malignant melanoma. J Clin Oncol. 2015;33:2500–2508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Song F, Qureshi AA, Giovannucci EL, et al. Risk of a second primary cancer after non-melanoma skin cancer in white men and women: A prospective cohort study. PLoS Med. 2013;10:e1001433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Colditz GA, Martin P, Stampfer MJ, et al. Validation of questionnaire information on risk factors and disease outcomes in a prospective cohort study of women. Am J Epidemiol. 1986;123:894–900. [DOI] [PubMed] [Google Scholar]
- 20. van Dam RM, Huang Z, Giovannucci E, et al. Diet and basal cell carcinoma of the skin in a prospective cohort of men. Am J Clin Nutr. 2000;71:135–141. [DOI] [PubMed] [Google Scholar]
- 21. Wu S, Han J, Laden F, Qureshi AA.. Long-term ultraviolet flux, other potential risk factors, and skin cancer risk: A cohort study. Cancer Epidemiol Biomarkers Prev. 2014;23:1080–1089. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Smith-Warner SA, Spiegelman D, Ritz J, et al. Methods for pooling results of epidemiologic studies: The Pooling Project of Prospective Studies of Diet and Cancer. Am J Epidemiol. 2006;163:1053–1064. [DOI] [PubMed] [Google Scholar]
- 23. Lee TK, MacArthur AC, Gallagher RP, Elwood MJ.. Occupational physical activity and risk of malignant melanoma: The Western Canada Melanoma Study. Melanoma Res. 2009;19:260–266. [DOI] [PubMed] [Google Scholar]
- 24. Thompson CA, Zhang ZF, Arah OA.. Competing risk bias to explain the inverse relationship between smoking and malignant melanoma. Eur J Epidemiol. 2013;28:557–567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Wu S, Han J, Li WQ, Li T, Qureshi AA.. Basal-cell carcinoma incidence and associated risk factors in U.S. women and men. Am J Epidemiol. 2013;178:890–897. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Fears TR, Bird CC, Guerry Dt, et al. Average midrange ultraviolet radiation flux and time outdoors predict melanoma risk. Cancer Res. 2002;62:3992–3996. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.