Does tattoo exposure increase the risk of cutaneous melanoma? A population-based case-control study
- Open Access
- 24.11.2025
- CANCER
Erschienen in:
Abstract
Introduction
The incidence of cutaneous melanoma has risen sharply over the past 30 years [1]. This has primarily been attributed to increased exposure to ultraviolet (UV) radiation [2], but chemical exposures are suspected to also play a role in the development of cutaneous melanoma, especially exposure to polyaromatic hydrocarbons (PAH) in occupational settings [3]. Risk factors related to lifestyle, such as tattoos, have been studied less. Historical tattoo prevalence data from Sweden are missing, but a Danish study of blood donors in 2001 showed that 10% (56/576) had a tattoo [4]. Recent estimates show that, in Sweden, 20% of the population is tattooed and the prevalence is above 30% in the population younger than 40 years [5]. The majority of people get their first tattoo at a young age – 18–35 years – and 20% are even younger [6]. This implies exposure to the inks for almost the entire life course, but little is known about the long-term health effects. Therefore, regulatory authorities have highlighted the need to clarify a potential link between tattoos and cancer.
Tattoo ink contains colour pigments – which are often the source of carcinogenic chemicals in tattoo ink – as well as solvents, preservatives, and impurities. Azo pigments dominate the organic colourants used in tattoo ink, which is problematic as they can be cleaved to carcinogenic aromatic amines, a process that is accelerated by UV radiation from exposure to sun, sunbeds or laser treatment [7]. Black ink often contains high amounts of PAH [8], some of which the International Agency for Research on Cancer [9] has classified as either carcinogenic or potentially carcinogenic, together with shading additives such as titanium dioxide [10]. Finally, heavy metals feature prominently irrespective of ink colour [11].
Anzeige
In January 2022, an EU regulation that imposes concentration limits for 4,000 chemicals in tattoo ink entered into force under Registration, Evaluation, Authorization and Restriction of Chemicals (REACH) [12]. Up until then, the market for tattoo ink had been relatively unregulated. Market controls by authorities show that inks contained toxic chemicals at levels above the stipulated limit values both before [13, 14] and after [15] the implementation of the EU regulation. The situation is often further complicated by fraudulent labelling of tattoo ink, where banned pigments are masked by listing legal pigments or label forging [13, 15].
An experimental study in mice has provided some evidence that red ink containing 2-anisidine may function as a weak cocarcinogen after irradiation with UV light [16]. However, no such studies have been undertaken to investigate the effect on CM onset.
Numerous case reports have described skin cancer in tattoos [17], but case reports suffer from inherent publication bias and cannot be used to infer causation. A recent Danish study of twins found an increased risk of skin cancer among tattooed individuals, but they only reported hazard ratios for melanoma and non-melanoma skin cancers combined [18]. Our aim was to answer if tattoo exposure increases the risk of cutaneous melanoma. Additionally, we hypothesized that exposure to UV radiation may act as an effect modifier, as it accelerates pigment breakdown into aromatic amines. We employed a population-based case-control design using national administrative registers and collected detailed exposure data using questionnaires.
Methods
Study design
In this population-based case-control study nested within the Swedish total population, all individuals with a diagnosis of nevus with severe atypia, cutaneous melanoma in situ or invasive cutaneous melanoma aged 20–60 years at diagnosis in 2017 (i.e. the index year) were identified in the National Cancer Register. Here, healthcare providers are obliged to report all primary neoplasms according to the Act on Health Data Registers (1998:543). The coverage of skin cancers is above 99% [19], and 99% of tumours in the skin are morphologically confirmed [20].
Anzeige
Three matched controls per case of the same age and sex were randomly sampled in the Total Population Register on the index date using incidence density sampling with person-time as the underlying timescale.
The study population was cross-referenced with the National Patient Register to exclude individuals at risk of suffering psychological distress due to study participation (i.e., F20-29 in the International Statistical Classification of Diseases and Related Health Problems [ICD], 10th Revision). If cases were deceased, the date of death was retrieved from the National Cause of Death Register. Next-of-kin were identified in the Total Population Register or the Multi-Generation Register and contacted to limit the impact of potential survivorship bias. Finally, indicator variables of socioeconomic status (SES) were retrieved from the Longitudinal Integration Database for Health Insurance and Labor Market Studies.
Exposure assessment
The survey was administered by Statistics Sweden and conducted between February and April 2021. We used a structured questionnaire to collect exposure data; it has been described in detail elsewhere [5].
In brief, we defined tattoos as permanent motifs obtained for decorative, cosmetic (i.e., permanent or semi-permanent makeup and microblading), or medical (i.e., in reconstruction of skin after surgery) reasons. Respondents were explicitly asked to also consider removed tattoos.
The study was powered to assess exposure as a dichotomous variable, i.e., as presence or absence of at least one tattoo. We assessed the area of the tattooed body surface as predefined increments on a categorical scale (i.e., < 1 palm, 1–5 palms, >5 palms), and as the number of undergone tattoo sessions.
We collected data on the respondents’ age when they obtained their first tattoo, and tattoo characteristics including colours, anatomical location, tattooist (professional or nonprofessional), and in which geographic region(s) the tattoos had been received.
The questionnaire contained validated items on recreational and occupational sun exposure [21], history of sunburns and use of tanning beds. In addition, the questionnaire asked about skin characteristics, i.e., pigmentation according to the Fitzpatrick Skin Phototype Classification type I-VI [22], and reaction to sun exposure derived from Fitzpatrick Skin Phototype Classification type I-IV. Finally, it also covered smoking.
Outcome
We defined incident primary cutaneous melanoma tumours according to ICD-O/3.2 (International Classification of Diseases for Oncology, 3rd Edition) typography code C44, in combination with any of the morphology codes listed in Table 1. In Sweden, dysplastic nevi (melanocytic nevi with severe atypia) are reported to the cancer registry. They are by definition not a cancerous lesion, but it is difficult to distinguish them clinically and often histopathologically from cutaneous melanoma in situ. In the present paper, we included them in the in situ/precursor lesion group of cutaneous melanomas because, in Sweden, that is how they are administered clinically. The main analysis of the present work refers to “overall cutaneous melanoma” and includes both precursor lesions and invasive cutaneous melanomas.
Anzeige
Table 1
International classification of diseases for Oncology, 3rd Edition, morphology codes (SNOMED; systematized nomenclature of Medicine) used in combination with typography code C44 to identify cases with cutaneous melanoma and registered associated diagnoses in the National cancer Register, with the number of cases (n) included in the present study
Morphology type | SNOMED | n |
|---|---|---|
In situ/precursor lesions | ||
Melanocytic nevus with severe atypia | 8727/0b | 717 |
Lentigo maligna | 8742/2b | 65 |
Acral lentiginous melanoma, in situ | 8744/2 | 4 |
Atypical Spitz’s nevus | 8770/1b | 9 |
Pigmented epithelioid melanocytoma/epithelioid blue nevus | 8780/1b | – |
Invasive cutaneous melanomas | ||
Superficial spreading melanoma | 8743/3 | 627 |
Nevoid melanoma and melanoma not otherwise specified | 8720/3 | 121 |
Lentigo malignant melanoma | 8742/3 | 18 |
Acral lentiginous melanoma | 8744/3 | 9 |
Nodular melanoma | 8721/3 | 25 |
Desmoplastic melanoma | 8745/3 | – |
Malignant melanoma in a giant pigmented nevus | 8761/3 | – |
Amelanotic melanoma | 8730/3 | – |
Mixed epithelioid and spindle cell melanoma | 8770/3 | 3 |
Melanoma from blue nevus | 8780/3 | – |
Study size
We determined the sample size through an a priori statistical power analysis considering tattoo status as a dichotomous variable. We wanted to be able to detect an odds ratio of 1.3 with 80% power, assuming a tattoo prevalence of 17% among controls [23]. This would be achieved by including 1300 cases and 2 controls per case. Acknowledging that questionnaires often suffer from low response rates, we scaled up the study population to 3000 cases and 3 controls per case to ensure sufficient power.
Statistical methods
For the questionnaire, an aggregated dropout analysis of characteristics of the participants and nonparticipants was performed by Statistics Sweden. Individual-level data from nonparticipants were not available.
To be considered exposed, participants were required to have obtained their first tattoo before the index year, to ascertain temporality between exposure and outcome.
Anzeige
For tattooed cases, we assessed the correlation between the tumour site and the anatomical location of tattoos using Spearman’s rho.
We used Directed Acyclic Graphs (DAG) in the design phase of the study to visualize potential confounders of the potential association between tattoo exposure and cutaneous melanoma (Figure S1). These assumptions informed the data collection strategy. We constructed an UV-exposure index to describe the cumulative UV exposure from numerous exposure sources according to the questionnaire items: “Occupational sun exposure (April-September),” “Recreational sun travel” and “Sunbed use.” Respondents were placed in the high category if they provided ≥ 1 response in the high exposure category on any of these items, in the medium category if they did not provide a response in any of the high categories and ≥ 1 response in a medium category, and in the low category if they only provided responses in the lowest categories.
We also created a phenotypic risk index to account for skin characteristics using the items skin pigmentation and skin reaction after sun exposure [22]. High-risk participants were those who reported “Skin reaction to first sun exposure”: “Burns (sometimes with blisters) and pain for at least 48 h” or “Burns with pain that disappears within 24 h,” regardless of their skin tone. Medium-risk participants were those who reported “Skin reaction to first sun exposure”: “First burn, then tan,” regardless of skin tone, and “Very fair” or “Fair” skin tone in combination with “Skin reaction to first sun exposure”: “Tan immediately.” Low-risk participants were those who reported “Medium,” “Olive,” “Brown” or “Dark brown” skin tone and reported “Skin reaction to first sun exposure”: “Tan immediately.”
We used logistic regression to investigate the association between tattoo exposure and cutaneous melanoma. The odds ratio offers an unbiased estimate of the incidence rate ratio (IRR) in the underlying population in case-control studies with incidence-density sampling. Exposure was modelled as a dichotomous variable: tattoo exposure (yes; no), as ordinal variables: size of tattoo (< 1 hand; >1 hand [1–5 hands and >5 hands were collapsed into the same category because overestimation of self-assessed tattoo size has been reported previously] [24]) and number of tattoo sessions (1; 2–3; ≥ 4 tattoo sessions), and as a nominal variable considering the colour scheme (black/grey; colour + black/grey; only colour).
Anzeige
We used conditional logistic regression for our primary analysis to align with the study protocol. In addition, we also ran unconditional logistic regression models where we broke the matching sets to optimize precision. Because we used balanced matching (case: control ratio 1:3), we accounted for the original confounding and the selection bias introduced by the matching by adjusting for sex, and for age (in 5-year intervals) with a term for residual age [25]. In the fully adjusted model, we further adjusted for household disposable income in the year before the index year (categorized according to quartiles according to the distribution among controls), marital status (married or registered partner; divorced or widowed; unmarried), educational level at the index year (primary and lower secondary; secondary; post-secondary), UV-exposure index (high; medium; low exposure), phenotypic risk index (high; medium; low risk), and smoking (current; previously; never) according to the DAG in Figure S2.
There were very few observations with missing data (1%) and we therefore restricted the analyses to observations with complete sets of covariates.
We explored the potential role of UV radiation and laser removal of tattoos as effect modifiers by introducing an interaction term between the dichotomous exposure variable and the UV-exposure index and constructing the exposure variable to distinguish tattooed individuals with respect to laser removal of tattoos, in separate models, respectively.
To clarify whether tattooed individuals had an increased risk of specific cutaneous melanoma subtypes, we performed subgroup analyses. First, we ran the analyses restricted to diagnostic codes for in situ/precursor lesions of cutaneous melanomas and invasive cutaneous melanomas, and then we restricted to melanocytic nevus with severe atypia and superficial spreading melanoma.
In sensitivity analyses, we first explored the impact of exposure duration by categorizing the exposure variable according to the number of years that had elapsed between the first tattoo and the index date (0–5; 5–10; 10–15; >15 years). Then, we performed a sensitivity analysis where we included sunburn during childhood (up to 13 years of age, many times; occasionally; never) as a covariate. Furthermore, to investigate the risk of CM in individuals with first ever diagnosis, we excluded those (n = 182) cases that had a previous CM diagnosis and performed analysis with only first ever diagnosis cases. To account for effects from immunosuppressive medication, we excluded individuals on such medications in another sensitivity analysis. Individuals with risk occupations associated with an increased risk for CM (firefighters and petroleum refinement workers) were excluded in a subsequent sensitivity analysis. Finally, we performed a sensitivity analysis where we included responses from deceased cases’ next-of-kin. This analysis could not be adjusted for phenotypic risk index, because next-of-kin received a reduced questionnaire that did not contain this information, and the data only allowed for an unmatched analysis.
Results
Study participants
Of the identified cases, 98% were still alive when the questionnaire was distributed (Fig. 1). The response rates among cases and controls were 56% and 46%, respectively. The drop-out analysis showed that, for both cases and controls, respondents were more often female, older, married, had higher educational attainment and income, and more often born in Sweden (Supplemental Table S1).
Fig. 1
Flow chart describing the inclusion of study participants. *Number of control respondents included in crude unmatched analysis were n = 4018. The remaining n = 79 controls were from risk sets with deceased cases and therefore only included in the next-of-kin sensitivity analysis
The median age at the index year was 51 (interquartile range = 43–56). Cases had slightly higher education level and disposable household income than controls, and they were more often married (Table 2). Overall, a higher proportion of cases were categorized as high risk according to the phenotypic risk index (Table 3). Regular recreational sun travel and use of sunbeds, as well as a higher proportion of sunburn during childhood were more frequent among cases, whereas controls were more likely to have experienced occupational sun exposure (Table 4).
Table 2
Characteristics of the participating cases and controls in the index year (2017)a
Cases n (%) | Controls n (%) | |
|---|---|---|
Tattoo status | ||
Yes | 354 (22) | 815 (20) |
No | 1244 (78) | 3203 (80) |
Missing | – | – |
Sex | ||
Male | 658 (41) | 1607 (40) |
Female | 940 (59) | 2411 (60) |
Age (years) | ||
< 29 | 64 (4) | 161 (4) |
30–39 | 191 (12) | 472 (12) |
40–49 | 450 (28) | 1137 (28) |
50–59 | 758 (47) | 1928 (48) |
60+ | 135 (8) | 320 (8) |
Missing | – | – |
Educational attainment | ||
Primary and lower secondary | 75 (5) | 267 (7) |
Upper secondary | 318 (20) | 973 (24) |
Post-secondary | 1203 (75) | 2768 (69) |
Missing | 2 | 10 |
Disposable income, household (SEK) | ||
< 330,900 | 297 (19) | 1003 (25) |
330,900–548,800 | 331 (21) | 1010 (25) |
548,900–738,000 | 437 (27) | 994 (25) |
≥ 738,100 | 532 (33) | 1010 (25) |
Missing | 1 | 1 |
Marital status | ||
Married/registered partnership | 918 (57) | 2168 (54) |
Divorced/widowed | 209 (13) | 552 (14) |
Unmarried | 471 (29) | 1298 (32) |
Missing | – | – |
Smoking | ||
Yes, current | 112 (7) | 432 (11) |
Yes, previously | 460 (29) | 1199 (30) |
No, never | 1024 (64) | 2376 (59) |
Missing | 2 | 11 |
Immunosuppressive medication | ||
Yes | 54 (3) | 90 (2) |
No | 1544 (97) | 3928 (98) |
Risk occupation | ||
Firefighter or Petrol refinement | 11 (1) | 29 (1) |
No risk occupation | 1587 (99) | 3989 (99) |
Table 3
Phenotypic characteristics of the cases and controls
Phenotype | Cases, n (%) | Controls, n (%) |
|---|---|---|
Eye colour | ||
Blue | 842 (53) | 1796 (45) |
Grey | 33 (2) | 100 (3) |
Green | 174 (11) | 419 (10) |
Yellow | 1 (0.06) | 5 (0.1) |
Light brown | 68 (4) | 252 (6) |
Dark brown | 95 (6) | 504 (13) |
Mixed (green, blue, grey) | 369 (23) | 891 (22) |
Different coloured eyes | 8 (0.5) | 27 (0.7) |
Missing | 8 | 24 |
Natural skin tone | ||
Very fair | 169 (11) | 229 (6) |
Fair | 940 (59) | 2007 (51) |
Medium | 463 (29) | 1590 (40) |
Olive/light brown | 10 (0.6) | 103 (3) |
Brown | 3 (0.2) | 23 (0.6) |
Dark brown | 0 (0) | 15 (0.4) |
Missing | 13 | 51 |
Freckles after sun exposure | ||
Yes, many | 260 (16) | 341 (9) |
Yes, a few | 604 (38) | 1272 (32) |
No, I don’t get freckles | 733 (46) | 2389 (60) |
Missing | 1 | 16 |
Skin reaction to first sun exposure | ||
Burns (sometimes with blisters) and pain for at least 48 h | 27 (2) | 57 (1) |
Burns with pain/sting that disappears within 24 h | 599 (38) | 1083 (27) |
Initially burns, then tans | 885 (56) | 2382 (60) |
Tans immediately | 79 (5) | 459 (12) |
Missing | 8 | 37 |
Phenotypic risk index | ||
High risk | 630 (39) | 1149 (29) |
Medium risk | 908 (57) | 2490 (62) |
Low risk | 59 (4) | 371 (9) |
Missing | 1 | 8 |
Table 4
Self-reported UV-exposure at the time of the survey
Cases, n (%) | Controls n (%) | |
|---|---|---|
Sunburn in childhood < 13 years of age | ||
Yes, many times | 240 (16) | 271 (7) |
Yes, occasionally | 650 (43) | 1187 (32) |
No, never | 622 (41) | 2291 (61) |
Missing | 86 | 269 |
Occupational sun exposure (April-September) | ||
Yes | 280 (18) | 835 (21) |
No | 1217 (76) | 2857 (71) |
I don’t work | 94 (6) | 306 (8) |
Missing | 7 | 20 |
Weekly hours of outside worka | ||
> 20 h/week | 71 (24) | 247 (28) |
6–20 h/week | 148 (49) | 382 (44) |
≤ 5 h/week | 80 (27) | 243 (28) |
Recreational sun travel > 13 years of age | ||
Yes, every year | 445 (28) | 922 (23) |
Yes, occasionally | 976 (62) | 2409 (61) |
No, never | 162 (10) | 648 (16) |
Missing | 15 | 39 |
Sunbed use > 13 years of age | ||
Several times a year | 403 (26) | 609 (15) |
Occasionally | 864 (54) | 2232 (56) |
Never | 321 (20) | 1144 (29) |
Missing | 10 | 33 |
UV-exposure index | ||
Ever high exposure | 736 (46) | 1531 (38) |
Never high exposure, ever medium exposure | 812 (51) | 2195 (55) |
Never high nor medium exposure | 47 (3) | 276 (7) |
Missing | – | – |
The overall tattoo prevalence was 21%. Among the cases, 354 respondents (22%) were tattooed before the index year, whereas the corresponding figure among controls was 815 (20%).
The median age for first tattoos was similar between cases (24 years) and controls (25 years, Table 5). Having a combination of black and coloured ink in tattoos was most common. Taken together, the variables estimating the total area the tattooed body surface indicated somewhat smaller tattoos among cases. It was most common to have a decorative tattoo, tattooed by a professional tattoo artist in Sweden. The proportions that were tattooed by a nonprofessional or outside Europe were similar between cases and controls.
Table 5
Exposure characteristics among tattooed cases (n = 354) and controls (n = 815)
Cases n (%) | Controls n (%) | |
|---|---|---|
Age at first tattoo (median, 25th −75th percentile) | 24 (18–35) | 25 (18–35) |
Type of tattooa | ||
Decorative | 338 (95) | 764 (94) |
Cosmetic | 29 (8) | 76 (9) |
Medical | 10 (3) | 15 (2) |
Missing | 20 | 41 |
Size | ||
> 5 palm | 35 (10) | 94 (12) |
1–5 palms | 126 (36) | 318 (39) |
< 1 palm | 191 (54) | 397 (49) |
Missing | 2 | 6 |
Colour | ||
Only black/grey | 149 (42) | 338 (42) |
Both black/grey and colour | 165 (47) | 384 (47) |
Only colour | 39 (11) | 90 (11) |
Missing | 1 | 3 |
Number of sessions | ||
1 | 143 (41) | 316 (39) |
2–3 | 123 (35) | 266 (33) |
4–5 | 35 (10) | 100 (12) |
6–9 | 26 (7) | 76 (9) |
>=10 | 24 (7) | 53 (7) |
Missing | 3 | 4 |
Tattooist | ||
Professional, studio | 319 (90) | 712 (87) |
Professional, other facility | 35 (10) | 93 (11) |
Cosmetic tattooist, in studio or clinic | 23 (7) | 59 (7) |
Healthcare professional, in clinic | 11 (3) | 12 (1) |
Nonprofessional, irrespective of location | 25 (7) | 64 (8) |
Missing | – | – |
Geographical regionb | ||
Sweden | 314 (89) | 715 (88) |
Nordics, except Sweden | 28 (8) | 49 (6) |
Rest of Europe | 33 (9) | 79 (10) |
Asia | 26 (7) | 55 (7) |
Oceania | 5 (1) | 5 (0.6) |
USA | 9 (3) | 18 (2) |
Other | 6 (2) | 25 (3) |
Missing | – | – |
Among the 270 (76%) of the tattooed cases, we had information about the anatomical location of the malignancy. In 30%, the tattoo was placed on the same anatomical location as the malignancy, although the correlation was low (Spearman’s rho = 0.007, p-value = 0.91).
Main analyses: tattoos and cutaneous melanoma
We found a 29% increased risk (IRR = 1.29, 95% CI: 1.07–1.56) of overall cutaneous melanoma among tattooed individuals in the matched adjusted analysis (Table 6). In the unmatched analysis, the risk was slightly attenuated (IRR = 1.24, 95% CI: 1.06–1.44). We did not find that larger tattoos resulted in a higher risk (Table 6). We found that individuals with tattoos that were both black/grey and coloured (IRR = 1.38, 95% CI: 1.08–1.77) had a higher risk compared to nontattooed individuals, and individuals with coloured tattoos also seemed to be at a similar risk, although the point estimate was associated with a wide confidence interval because of small numbers. We found a consistent risk increase over all exposure duration intervals, except 5–10 years where the estimate was null (Table 6). Both shorter exposure duration 0–5 years and a longer exposure duration (10–15 years) were associated with increased risks, IRR = 1.60 (95% CI: 0.99–2.58) and IRR = 1.67 (95% CI: 1.09–2.56), respectively. Individuals with over 15 years of tattoo exposure had a slightly lower risk increase (IRR = 1.24, 95% CI: 0.98–1.57). In contrast, the unmatched analysis showed a consistent risk increase over all exposure durations. We did not find evidence of effect modification by UV exposure (P[interaction] = 0.58. Table S2). Adding childhood sunburn to the model increased the effect estimate to 1.31 (95% CI 1.07–1.61) (Table 6). Very few participants had removed tattoos using laser, and their effect estimates were below 1 and associated with a very wide confidence interval (Table 6).
We found increased risks for both subgroups of cutaneous melanoma; invasive cutaneous melanomas (IRR = 1.25, 95% CI = 0.95–1.64) and in situ cutaneous melanomas (IRR = 1.33, 95% CI = 1.02–1.75) (Fig. 2, Table S3). When analysing the risk for the single largest type in each subgroup, we found even higher risk estimates; superficial spreading melanoma (IRR = 1.40, 95% CI = 1.03–1.90) and melanocytic nevus with severe atypia (IRR = 1.39, 95% CI = 1.05–1.85). An exposure duration of 10 to 15 years was associated with the highest point estimates irrespective of cutaneous melanoma subtype.
When excluding individuals who had a previous CM diagnosis (Table S4), we found increased risks for first CM diagnosis cases (IRR = 1.23, 95% CI = 1.01–1.51), however lower than the main results. When excluding individuals with immunosuppressive medications, we found that the risk increased further (IRR = 1.32, 95% CI = 1.10–1.62). When excluding individuals with risk occupations risk estimates did not deviate from the main result.
Including the responses from the 11 participating next-of-kin did not affect the results (Table S5).
Fig. 2
Adjusted incidence rate ratios (IRR) of cutaneous melanoma subgroups relative to non-tattooed participants. Tattoo exposure was modelled as exposure status (tattooed vs. nontattooed), and exposure duration (0-5[red], 5-10[green], 10-15[brown], >15 years[purple] vs. non-tattooed)
Table 6
Incidence rate ratios (IRR) of cutaneous melanoma in tattooed individuals relative to non-tattooed individuals
Matched analysis | Unmatched analysis | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Basic adjustmenta | Full adjustmentb | Basic adjustmentc | Full adjustmentd | |||||||||
Cases (n) | Controls (n) | IRRe (95% CI) | Cases (n) | Controls (n) | IRR (95% CI) | Cases (n) | Controls (n) | IRR (95% CI) | Cases (n) | Controls (n) | IRR (95% CI) | |
Tattoo status (main analysis) | 1333 | 2271 | 1328 | 2243 | 1598 | 4018 | 1592 | 3979 | ||||
Tattooed | 285 | 437 | 1.13 (0.95–1.34) | 284 | 432 | 1.29 (1.07–1.56) | 354 | 815 | 1.13 (0.98–1.30) | 353 | 809 | 1.24 (1.06–1.44) |
Non-tattooed | 1048 | 1834 | 1.00 | 1044 | 1811 | 1.00 | 1244 | 3203 | 1.00 | 1239 | 3170 | 1.00 |
Tattooed body surface | 1331 | 2268 | 1327 | 2240 | 1596 | 4012 | 1591 | 3973 | ||||
Tattooed body surface < 1 palm | 159 | 216 | 1.31 (1.05–1.68) | 159 | 213 | 1.47 (1.16–1.87) | 191 | 397 | 1.25 (1.04–1.51) | 191 | 394 | 1.33 (1.10–1.61) |
Tattooed body surface > 1 palms | 124 | 218 | 0.95 (0.75–1.22) | 124 | 216 | 1.12 (0.86–1.45) | 161 | 412 | 1.01 (0.83–1.23) | 161 | 409 | 1.15 (0.93–1.42) |
Non-tattooed | 1048 | 1834 | 1.00 | 1044 | 1811 | 1.00 | 1244 | 3203 | 1.00 | 1239 | 3170 | 1.00 |
Ordinal exposure variable | 1330 | 2268 | 1326 | 2240 | 1595 | 4014 | 1590 | 3975 | ||||
1 tattoo sessions | 121 | 178 | 1.19 (0.93–1.53) | 121 | 173 | 1.32 (1.01–1.73) | 143 | 316 | 1.17 (0.95–1.45) | 143 | 311 | 1.28 (1.03–1.59) |
2–3 tattoo sessions | 96 | 142 | 1.17 (0.89–1.55) | 96 | 142 | 1.37 (1.02–1.84) | 123 | 266 | 1.20 (0.96–1.50) | 123 | 266 | 1.29 (1.03–1.63) |
>4 tattoo sessions | 65 | 114 | 0.96 (0.69–1.34) | 65 | 114 | 1.14 (0.81–1.61) | 85 | 229 | 0.96 (0.74–1.25) | 85 | 228 | 1.10 (0.83–1.44) |
Non-tattooed | 1048 | 1834 | 1.00 | 1044 | 1811 | 1.00 | 1244 | 3203 | 1.00 | 1239 | 3170 | 1.00 |
Colour | 1332 | 2269 | 1327 | 2241 | 1597 | 4015 | 1591 | 3976 | ||||
Black/grey | 111 | 180 | 1.04 (0.80–1.34) | 110 | 178 | 1.17 (0.88–1.54) | 149 | 338 | 1.15 (0.94–1.42) | 148 | 335 | 1.23 (0.99–1.53) |
Black/grey + colour | 142 | 209 | 1.17 (0.93–1.47) | 142 | 209 | 1.38 (1.08–1.77) | 165 | 384 | 1.11 (0.92–1.35) | 165 | 384 | 1.27 (1.04–1.56) |
Colour only | 31 | 46 | 1.31 (0.82–2.11) | 31 | 43 | 1.40 (0.85–2.31) | 39 | 90 | 1.11 (0.76–1.63) | 39 | 87 | 1.17 (0.79–1.73) |
Non-tattooed | 1048 | 1834 | 1.00 | 1044 | 1811 | 1.00 | 1244 | 3203 | 1.00 | 1239 | 3170 | 1.00 |
Exposure duration | 1333 | 2271 | 1328 | 2243 | 1598 | 4018 | 1592 | 3979 | ||||
0–5 years | 38 | 48 | 1.44 (0.91–2.28) | 38 | 47 | 1.60 (0.99–2.58) | 45 | 98 | 1.19 (0.82–1.72) | 45 | 97 | 1.25 (0.85–1.83) |
5–10 years | 37 | 70 | 0.92 (0.61–1.40) | 36 | 70 | 0.98 (0.63–1.53) | 55 | 125 | 1.17 (0.84–1.64) | 54 | 125 | 1.27 (0.90–1.79) |
10–15 years | 49 | 55 | 1.50 (1.00–2.25.00.25 | 49 | 55 | 1.67 (1.09–2.56) | 62 | 110 | 1.50 (1.08–2.07) | 62 | 110 | 1.60 (1.15–2.24) |
Over 15 years | 161 | 264 | 1.06 (0.85–1.31) | 161 | 260 | 1.24 (0.98–1.57) | 192 | 482 | 1.03 (0.86–1.23) | 192 | 477 | 1.15 (0.95–1.39) |
Non-tattooed | 1048 | 1834 | 1.00 | 1044 | 1811 | 1.00 | 1244 | 3203 | 1.00 | 1239 | 3170 | 1.00 |
Childhood sunburnf | 1333 | 2271 | 1252 | 2098 | 1598 | 4018 | 1507 | 3722 | ||||
Tattooed | 285 | 437 | 1.13 (0.95–1.34) | 270 | 403 | 1.31 (1.07–1.61) | 354 | 815 | 1.13 (0.98–1.30) | 334 | 748 | 1.23 (1.05–1.44) |
Non-tattooed | 1048 | 1834 | 1.00 | 982 | 1695 | 1.00 | 1244 | 3203 | 1.00 | 1173 | 2974 | 1.00 |
Laser removal | 1331 | 2268 | 1327 | 2240 | 1596 | 4012 | 1591 | 3974 | ||||
Tattooed, laser removal | 5 | 16 | 0.44 (0.16–1.24) | 5 | 16 | 0.64 (0.22–1.82) | 7 | 25 | 0.73 (0.31–1.69) | 7 | 25 | 0.85 (0.36–2.01) |
Tattooed, no laser removal | 278 | 418 | 1.16 (0.97–1.38) | 278 | 413 | 1.32 (1.09–1.60) | 345 | 784 | 1.14 (0.99–1.32) | 345 | 779 | 1.25 (1.08–1.46) |
Non-tattooed | 1048 | 1834 | 1.00 | 1044 | 1811 | 1.00 | 1244 | 3203 | 1.00 | 1239 | 3170 | 1.00 |
Discussion
The results suggested that tattooed individuals had a 29% increased risk of cutaneous melanoma compared to nontattooed individuals. When only cases with a first diagnosis were included, tattooed individuals had a 23% increased risk for cutaneous melanoma. According to various studies and meta-analyses, the risk associated with UV radiation—the most well-established risk factor of cutaneous melanoma—is estimated to be comparable to up to five times higher [26‐30] than the risk observed in our study. If tattoo exposure is an additional risk factor for cutaneous melanoma, it may exacerbate the risk imposed by UV exposure when tattooed skin is exposed to the sun.
We did not find that highly UV-exposed tattooed individuals had a higher risk of cutaneous melanoma than tattooed individuals with lower UV exposure. Thus, our data suggests that UV-accelerated pigment breakdown does not increase the risk of cutaneous melanoma beyond the risk associated with tattoo exposure per se, but further studies in more UV-intense regions of the world is warranted.
In 2012, a literature review of case reports [17] concluded that the link between tattoos and skin cancer was likely coincidental. The authors reached this conclusion by comparing the number of published case reports of cutaneous malignancies in tattoos with the size of the estimated global tattooed population. However, case studies suffer from inherent publication bias due to selective reporting and cannot be used to evaluate associations between exposures and outcomes. Thus, our results provide the first insight into a possible association between tattoos and skin cancer using an adequate methodological approach.
We did not see a higher risk increase for those with larger areas of tattooed skin, rather their IRRs were lower than those of individuals with smaller tattoos. It is possible there was misclassification in the ordinal exposure variable because tattooed individuals tend to overestimate their tattooed body area [24]. It could also be caused by variation in the amount of ink needed for tattooing [31]. In the matched analysis, coloured tattoos alone or in combination with black or grey in the tattoo seemed to be associated with a slightly higher risk of cutaneous melanoma. However, statistical power for the group with only coloured tattoos was an issue, and further investigation is warranted to clarify if there is a difference in cancer risk depending on the ink colour.
Depending on the mechanism of carcinogenicity, chemicals can act as tumour initiators or tumour promotors [32], either initiating the cancerous development of a cell – which would indicate a longer latency period between exposure and diagnosis – or promoting or progressing its development resulting in a shorter latency period. The biological mechanism underlying cancer development should inform the analytical choices regarding latency times but given the complex mixture of chemicals in tattoo ink, this was not possible. Instead, we chose to adjust for exposure duration. We found that those with an exposure duration of 10–15 years were at highest risk of developing cutaneous melanoma. The same time span is generally implicated in UV-induced cutaneous melanoma. We also observed high risk estimates in individuals with short exposure duration (up to five years), but the confidence interval was wide. Mechanistically, the results suggest that tattoo exposure could be associated with tumour initiation, and possibly with tumour promotion, but future research is needed to confirm this for example by longitudinal studies following tattooed individuals.
In our previous study on the association of cutaneous squamous cell carcinoma (CSCC) and tattoos, using the same study design as in the present study, we did not find an increased risk for tattooed individuals [33]. Nielsen et al. [34] found an increased risk for lymphoma for tattooed individuals, also using the same study design as in the present study, and Clemmensen et al. [18] found evidence for an increased risk for tattooed individuals for lymphoma and skin cancers in a Danish twin study. There now are three epidemiological studies suggesting an association between tattoos and cancer – lymphoma, and cutaneous melanoma. These are both cancers with a strong immunologic component, while SCC is not. Cutaneous melanoma is a highly immunogenic cancer with high infiltrates of immune cells, and sophisticated evasion tactics, that respond well to immunotherapy [35‐38]. Squamous cell carcinoma, while still influenced by the immune system, is less immunogenic, with less immune infiltrates and has less complex immune evasion tactics, that responds less well to immunotherapy [39, 40]. We propose that carcinogenic mechanisms may stem from immunotoxicity of tattoo ink. Studies using reconstructed human skin show that tattoo ink can present immunotoxic properties such as cytotoxicity and act as sensitizers [41, 42], which might provide a pathobiological mechanism to explain the epidemiologic results, and why we see an increased risk for CM and not for cSCC. Our argument is further supported by the stronger association observed when we exclude participants on immunosuppressive medications. This exclusion results in an even stronger association and a 32% increase in risk.
In the subgroup analysis, we found that the risk associated with tattoo exposure seemed to be increased for both main groups: invasive and in-situ cutaneous melanomas, attributable to both superficial spreading melanoma and melanocytic nevus with severe atypia. Tattooed individuals had an approximate risk increase of 40% compared to nontattooed individuals for both subtypes. Superficially spreading melanomas constitute an invasive cancer, which often shows histopathological remnants of melanocytic nevi in part of the lesion [43]. Therefore, it may be advisable to take extra precautions when clinically evaluating nevi and the skin in tattooed individuals. After excluding cases with previous diagnosis of CM, the risk increase was lower (23%) than for the entire group (29%). Cases with multiple diagnosis of CM is likely a group with many life-style related and genetic risk factors, which may affect the risk estimate, as well as decreased power when excluding 182 individuals and their matched controls, which could explain the slightly lower risk increase of 23%.
Tattoo-related factors, other than the pigments as such, that could theoretically play a role in the carcinogenesis are the trauma induced by puncturing the skin, the scarring process, and exposure to solar irradiation when tattoos are not covered with clothing [17]. Biomolecular changes indicative of chronic inflammation have been described in skin adjacent to tattoo pigments [44], and the role of inflammation in the development of cancer is well established [45]. Only 30% of the tattoos were located on the same anatomical site as the malignancy. This may seem illogical, but studies of firefighters [46] and petroleum workers [47, 48] have demonstrated increased risks of cutaneous melanomas despite the fact that their exposure to chemicals (including, e.g., PAH) usually is more systemic, through uptake via lungs and skin [49, 50], and not necessarily localized to the anatomical location of the malignancy. Another study of Danish twins found the same association, although for both melanoma and non-melanoma skin cancers [18].
It is not clear for how long exposure to soluble ink components can be expected systemically, but systemic exposure may not be necessary. Instead, the possible effects that components of tattoo ink could exert on the immune system locally (i.e. at the site of the tattoo, and in the proximate lymph nodes) may become systemic. However, the toxicokinetics of tattoo ink exposure is an understudied area, and a proper risk assessment can be made once more is known about the kinetics and dynamics of tattoo ink in the body.
Strengths and limitations
Our study is the first epidemiological study to investigate the association between tattoos and cutaneous melanoma using a large population-based sample with access to information about important confounders. We consider the exposure variable – having a tattoo – to be non-sensitive to recall bias owing to the invasive nature of tattoos. Importantly, we found that common risk factors for cutaneous melanoma, such as high UV exposure and light skin, were more common among cases than controls, which suggests that our results are valid. Given the response rate of 50% in the present study, selection bias could have been introduced if the likelihood of responding was related to both the exposure and the outcome. However, to decrease the risk of conditioning participation on exposure or outcome status, the questionnaire asked about several lifestyle-related factors and informed participants that the purpose of the study was to investigate the association between lifestyle-related factors and cancer as well as other diseases.
A limitation may be that responders and non-responders differed with respect to SES. Because we assume that UV exposure is dependent on SES, selection bias with respect to SES could have affected the estimates, and particularly the lack of interaction we found for tattoos and UV-exposure. The results may not be generalizable to older age groups, given the different presentation, clinical course and prognosis of cutaneous melanoma [51]. Caution should be taken in generalizing our results to a setting with higher UV exposure, because we could only investigate the potential effect modification in a setting with relatively low UV exposure.
Childhood sun exposure is a known risk factor for CM [27, 52], and in the context of our investigation on CM risk for tattooed individuals, it is likely an important precision variable. The item reflecting childhood sun exposure relevant to our investigation was childhood sunburn. Our risk and sun indexes did not include childhood sunburn, because we wanted them to reflect the participants’ risk behaviour in adult life. We also decided not to include childhood sunburn in the main analysis, because of the higher missingness in the variable. However, after adding childhood sunburn to a sensitivity analysis, we found that this increased the risk for tattooed individuals further, and we may therefore underestimate the CM risk for tattooed individuals in our main analysis.
Our data included mainly melanocytic nevi with severe atypia and superficial spreading melanoma. Therefore, we cannot draw conclusions as to the association between tattoo exposure and the less common types of cutaneous melanomas, because of limited statistical power.
Conclusions
The results suggested that tattoos may be a risk factor for cutaneous melanoma. The results need to be verified in additional epidemiologic studies before causality can be inferred.
Acknowledgements
The authors thank Professor Thomas Rustemeyer for his valuable input on the Discussion.
Declarations
Competing interests
The work was supported by research funding to CN from the Swedish Research Council for Health, Working Life and Welfare, The Crafoord Foundation, and Magnus Bergvalls Foundation, to KN from the Southern health care region, Sweden, the S.R. Gorthon foundation, the Welander Finsen foundation (Hudfonden), the LMK foundation, grants from the Swedish state under the agreement between the Swedish government and the county councils (the ALF agreement), the Inga and John Hain Foundation for medical research, and the Gyllenstiernska Krapperup Foundation. KN has received speaker honoraria from Galderma Sweden, LEO Pharma Sweden, Novartis Sweden and UCB Pharma Sweden. KN has 2021 served on one advisory board for Merck Sharp and Dohme (MSD), and finally KN is President of the Board for the National Melanoma Guidelines and President for the Swedish Society of Dermatological Surgery and Oncology. The authors declare no other relationships or activities that could appear to have influenced the submitted work.
Ethical approval and consent to participate
The study was approved by the Swedish Ethical Review Authority (no. 2019–03138). Participants consented to participation by answering the questionnaire. The study was performed in accordance with the Declaration of Helsinki.
Consent for publication
No personal data was published on an individual level in this study.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.