Abstract
Background
Several observational studies have suggested differences in the risk factor profile between patients with superficial basal cell carcinomas (BCCs) and non‐superficial BCCs.
Objective
To test the reproducibility of previous study findings and to find new genetic and non‐genetic predictors for patients with a superficial first BCC.
Methods
A total of 14.628 participants of northwestern European descent aged 45 years or older from a prospective population‐based cohort study (Rotterdam Study) were linked with the Dutch Pathology Registry (PALGA) of whom 1528 were identified as BCC patients. After exclusion, 948 eligible BCC patients remained for further non‐genetic analyses and 1014 for genetic analyses. We included 11 phenotypic, environmental and tumour‐specific characteristics, and 20 candidate single nucleotide polymorphisms (SNP) as potential predictors for patients with a superficial first BCC. We performed binary logistic multivariable regression analyses.
Results
We found that patients with a superficial first BCC were significantly younger, almost two times more often female and 12–18 times more likely to have their BCC on the trunk or extremities than patients with a non‐superficial first BCC. One SNP (rs12203592), mapped to IRF4, looked promising (OR 1.83, 95% CI 1.13–2.97, P‐value <0.05), but after adjustment for multiple testing, no significant differences in genetic make‐up between superficial BCC and non‐superficial BCC patients were found.
Conclusion
We conclude that patients with a superficial BCC differ from non‐superficial BCC patients with respect to environmental factors (tumour localization as a proxy for UVR exposure) and phenotypic characteristics (age and sex), but we found no difference in genotype. As superficial BCC patients develop their first BCCs at a younger age, they could be at higher lifetime risk for subsequent skin cancers and therefore be an important group for secondary prevention.
Short abstract
Linked article: This article is commented on G. Argenziano et. al., p. 461 in this issue. To view this article visit https://doi.org/10.1111/jdv.15486.
Introduction
Patients with basal cell carcinoma (BCC) put a strain on healthcare services in countries with mainly white‐skinned inhabitants, as a result of the high and increasing BCC incidence, and the increased risk of synchronous and metachronous BCCs and other ultraviolet radiation (UVR)‐related skin cancers (i.e. field cancerization).1, 2, 3 In addition, the disability‐adjusted life years and healthcare costs for BCC have risen significantly as well.4, 5
There are different histopathological subtypes of BCC, based on the growth pattern(s) found within the tumour tissue. The nodular pattern is the most frequently found histological subtype (>50%), followed by superficial (~20%) and infiltrative (~10%), and about 20% of the tumours show a mixed type.6, 7, 8, 9, 10, 11 The frequencies reported depend on the used pathological classification system and period of the study, because classification systems and subtype incidences changed over time.12, 13 BCCs mostly occur on the head and neck area (i.e. chronically sun exposed; >70%), followed by the trunk (~20%) and extremities (~10%), which are both areas intermittently exposed to UVR.7, 8, 9, 10, 14 Several observational studies have identified associations between age, sex and anatomical site, and BCC subtypes.6, 7, 8, 9, 13 Patients with a superficial BCC more often have their BCC on the trunk and extremities than in the head and neck region,6, 7, 8, 9, 13 are younger7, 8, 9, 13 and more often female.8, 9 In addition, patients with an initial truncal superficial BCC developed metachronous BCCs at a faster rate than patients with other anatomical site and histology combinations.15
These results could indicate that different BCC subtypes, in particular superficial, have other aetiologies with respect to environmental factors (e.g. UVR exposure), phenotypic characteristics (e.g. age and sex) and genetic predisposition. However, only a few studies have studied other predictors than age, sex and anatomical site, with conflicting results.10, 16, 17
The objective of this study was to test the reproducibility of these findings and to find potentially new predictors for patients with a superficial first BCC (sBCC). We hereto analysed the data of almost 1000 white‐skinned participants with a BCC of a prospective population‐based cohort study (Rotterdam Study).
Materials and methods
Study population
The Rotterdam Study is a prospective population‐based cohort study of 14 926 participants (divided over three cohorts) aged 45 years or older, living in a well‐defined suburb of Rotterdam, the Netherlands.18 The cohorts predominantly consist of people of northwestern European descent. All the participants were interviewed and examined at baseline, and these examinations were repeated about every 4 years. The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus MC and by the Ministry of Health, Welfare and Sport of the Netherlands, implementing the Wet Bevolkingsonderzoek: ERGO (Population Studies Act: Rotterdam Study) and it was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent to participate in the study and to obtain information from their treating physicians.
Phenotype/Case definition
The method by which we identified BCCs has been described in detail previously.19 In short, the study database was linked to the Dutch nationwide network and registry of histopathology and cytopathology (PALGA) to retrieve medical history of all participants on histopathologically confirmed BCCs between 1 July 1989 and 31 December 2013.20 Of the 14 926 Rotterdam Study participants, 298 did not sign informed consent for a linkage and could not be linked to PALGA. The pathology excerpts we received contained information on date of diagnosis, anatomical location, body side, type of procedure (i.e. biopsy or excision), radicality and diagnosis. The majority of these excerpts showed a subtyping of the BCC, and these subtypes were coded based on the World Health Organization's histological classification of keratinocytic skin tumours.21 If there was a subtype discrepancy between a biopsy and an excision or a biopsy/excision included more than one subtype, we coded it as a mixed type BCC and noted the concerned subtypes. Patients with a missing subtype were excluded and patients with a mixed type first BCC with a superficial component were excluded as well, because it was unclear to which subtype these belong (i.e. superficial or non‐superficial). Metachronous BCCs that occurred within 6 months of the first BCC were counted as additional tumours at the date of the initial diagnosis, as those BCCs were most likely present at this earlier date. We randomly selected a BCC for participants with synchronous BCCs on their first diagnosis date.
Selection of non‐genetic candidate predictors
A literature search up to May 2016 for English publications on phenotypic, environmental and tumour‐specific factors previously involved in BCC subtypes was done in PubMed. Four phenotypic factors were included, namely age at first BCC, sex, tendency to develop sunburn and pigment status.7, 8, 9, 10, 13 The latter was a combination of eye colour and hair colour when young (e.g. a participant with blue eyes and red hair was scored as light). Five environmental characteristics were chosen and concerned a history of being outdoor for over 4 h per day during more than 25 years, sun protective behaviour measured by wearing sunglasses or a hat, smoking, alcohol consumption and coffee consumption.10, 22, 23, 24 Finally, two tumour‐related variables were included, namely localization of the first BCC and the number of BCCs at first date of diagnosis.6, 7, 8, 9, 13, 15 All selected variables (except tumour‐specific characteristics) were measured at study entry or at a study visit closest to study entry.
Selection of candidate single nucleotide polymorphisms
A literature search up to May 2016 for English genome‐wide association study (GWAS) publications of loci that confer risk of BCC or non‐melanoma skin cancer was done in PubMed. There was no GWAS of the histopathological subtypes of BCC. To reduce the burden of multiple testing, all selected single nucleotide polymorphisms (SNPs) had to be at least borderline genome‐wide significant (P‐value <7.0 × 10−8) and had to be replicated in another cohort. This resulted in a list of 20 candidate SNPs located in 17 different chromosomal regions (Table S1).25
Genotype
DNA was isolated from whole blood, further processed and quality checked following standard protocols.18 The Illumina Infinium II HumanHap550 BeadChips and the Illumina Human610‐Quad BeadChips were used to genotype the Rotterdam Study participants.
Quality control criteria included removing SNPs with Hardy–Weinberg equilibrium deviations (P‐value <0.0001), genotyping call rate <97%, gender mismatch and a high mean autosomal heterozygosity. SNPs were not included if they had a minor allele frequency of <1% and/or an imputation r² of <0.3.
For the candidate SNP approach, we used genotypes that were estimated from the imputed 1000 Genomes, GIANT Phase I version 3 dosage data18 using the Genome‐wide Complex Trait Analysis (GCTA) software with default parameters.26 All selected candidate SNPs were included in our genetic database.
Statistical analysis
Non‐genetic binary logistic regression analysis of sBCC vs. non‐superficial BCC (nsBCC)
All the assumptions of a binary logistic regression analysis were tested and we found no violations. There existed no strong (multi)collinearity between the selected non‐genetic candidate predictors. A few outliers in the coffee consumption and alcohol consumption variables were found using the outlier labelling rule,27 but all values were realistic. There was sufficient power to include the 11 selected candidate predictors in the multivariable binary logistic regression analysis.
We could safely assume that missing predictor values were missing at random (i.e. missing data points were not related to the missing data itself, but to the observed data). Missing predictor values could therefore be imputed using multiple imputations (30 times) by an iterative Markov Chain Monte Carlo method. The imputation model included all candidate predictors, the outcome, the body mass index (kg/m2), the level of education, the side of the first BCC and the Rotterdam Study cohort number. After the imputations, we did both univariable and multivariable binary logistic regression analyses. No selection methods were used for the multivariable analysis.
All of the data management and the non‐genetic binary logistic regression analyses were done in IBM® SPSS® Statistics for Windows version 21 (Chicago, IL, USA).
Genetic (SNP‐based) binary logistic regression analysis of sBCC vs. nsBCC
All the assumptions of a binary logistic regression analysis were tested and we found one violation, namely collinearity between two selected candidate SNPs. A bivariate correlation matrix showed a Pearson correlation coefficient of 0.86 between rs12210050 and rs12202284, which means that these predictors were highly correlated. A few outliers in the age and principal component variables were found using the outlier labelling rule,27 but all values were realistic. There was insufficient power to include the 20 selected candidate SNPs, age, sex and four principal components (PCs) in the multivariable binary logistic regression analysis. Therefore, we adjusted our analyses for multiple testing using the false discovery rate (FDR).28 PCs were included to adjust for possible population stratification.
The SNP‐based association analyses were performed on the imputed dosage data using a binary logistic regression with an additive model. The multivariable logistic regression analysis was adjusted for age at BCC diagnosis, sex and four PCs. No selection methods were used for the multivariable analysis.
The genetic data were prepared on our genetic servers, and IBM® SPSS® Statistics for Windows version 21 was used for the analyses.
Sensitivity analyses of sBCC vs. nodular BCC
We performed sensitivity analyses by doing the same non‐genetic and genetic regression analyses as for sBCC vs. nsBCC, but now including only patients with superficial or nodular first BCC.
Results
Study population for non‐genetic analyses
Of the 14 628 Rotterdam Study participants linked to PALGA, 1528 had at least one BCC. After the exclusion of patients with a missing subtype (n = 71), patients with a mixed superficial first BCC (n = 58) and patients who developed at least one BCC before study entry (n = 451), 948 eligible BCC patients remained for further analyses. We randomly selected a BCC for participants with synchronous BCCs on their first diagnosis date (n = 125). Of the included patients, 137 (14%) had a superficial first BCC, 496 (52%) a nodular first BCC and the remaining 315 (33%) another subtype (infiltrative, micronodular or non‐superficial mixed type; Tables 1 and S2).
Table 1.
Non‐genetic characteristics of 948 Rotterdam Study patients with a first BCC
| Patient and tumour characteristics | Coding | Overalla | Superficial BCC | Non‐superficial BCC |
|---|---|---|---|---|
| Number of patients | 948 (100%) | 137 (100%) | 811 (100%) | |
| Age at first BCC (years) | Median (IQR) | 74.6 (67.9–81.2) | 70.2 (64.3–76.0) | 75.5 (68.9–81.8) |
| Sex | Female | 526 (55%) | 87 (64%) | 439 (54%) |
| Pigment status | Dark | 153 (16%) | 24 (18%) | 129 (16%) |
| Intermediate | 447 (47%) | 70 (51%) | 377 (46%) | |
| Light | 213 (22%) | 30 (22%) | 183 (23%) | |
| Missing | 135 (14%) | 13 (9%) | 122 (15%) | |
| Easily sunburned | Yes | 319 (34%) | 53 (39%) | 266 (33%) |
| Missing | 65 (7%) | 5 (4%) | 60 (7%) | |
| Outdoor work | Yes | 124 (13%) | 14 (10%) | 110 (14%) |
| Missing | 274 (29%) | 42 (31%) | 232 (29%) | |
| Sun protection | No, never or hardly ever | 357 (38%) | 44 (32%) | 313 (39%) |
| Missing | 60 (6%) | 4 (3%) | 56 (7%) | |
| Smoking | Current or former | 623 (66%) | 92 (67%) | 531 (65%) |
| Missing | 17 (2%) | 1 (1%) | 16 (2%) | |
| Alcohol consumption (glasses/day) | Median (IQR) | 0.6 (0.1–1.7) | 0.6 (0.1–1.4) | 0.6 (0.1–1.8) |
| Missing | 215 (23%) | 18 (13%) | 197 (24%) | |
| Coffee consumption (cups/day) | Median (IQR) | 3.3 (2.0–4.5) | 3.3 (1.5–4.0) | 4.0 (2.0–5.0) |
| Missing | 215 (23%) | 18 (13%) | 197 (24%) | |
| >1 BCC at initial diagnosis | Yes | 125 (13%) | 24 (18%) | 101 (12%) |
| Localization of first BCC | Head and neck | 630 (66%) | 24 (18%) | 606 (75%) |
| Extremities | 128 (14%) | 54 (39%) | 74 (9%) | |
| Trunk | 184 (19%) | 58 (42%) | 126 (16%) | |
| Missing | 6 (1%) | 1 (1%) | 5 (1%) |
Participants with a mixed type BCC with a superficial component were excluded.
BCC, basal cell carcinoma; IQR, interquartile range.
Patients with a superficial first BCC were younger than patients with a non‐superficial first BCC (median age 70.2 vs. 75.5 years), and the proportion females (64%) were higher in sBCC patients than in nsBCC patients (54%; Table 1). Approximately 4 out of 5 sBCCs were located on the extremities (39%) or trunk (42%) as opposed to 1 in 4 of the nsBCCs.
Non‐genetic binary logistic regression analyses of sBCC vs. nsBCC
Of the 11 candidate predictors, 3 were significantly associated with a superficial first BCC in the univariable binary logistic regression analyses, namely a younger age at first BCC diagnosis (OR: 0.94, 95% CI: 0.92–0.96 per year), female gender (OR: 1.47, 95% CI: 1.01–2.14) and localization on the trunk (OR: 11.44, 95% CI: 6.85–19.10) or extremities (OR: 18.07, 95% CI: 10.56–30.93; Table 2).
Table 2.
Associations between non‐genetic predictors and occurrence of superficial first BCC (n = 948)†
| Patient and tumour characteristics | Coding | Univariable models‡ | Multivariable model‡ , § |
|---|---|---|---|
| Age at first BCC (years) | Continuous | 0.94 (0.92–0.96)*** | 0.95 (0.93–0.98)*** |
| Sex | Female | 1.47 (1.01–2.14)* | 1.88 (1.16–3.03)* |
| Pigment status | Dark | Reference | Reference |
| Intermediate | 1.01 (0.61–1.67) | 0.91 (0.50–1.64) | |
| Light | 0.92 (0.51–1.65) | 0.80 (0.40–1.61) | |
| Easily sunburned | Yes | 1.22 (0.83–1.78) | 1.13 (0.70–1.81) |
| Outdoor work | Yes | 0.77 (0.42–1.38) | 0.85 (0.43–1.69) |
| Sun protection | No or hardly ever | 0.70 (0.48–1.04) | 0.80 (0.51–1.26) |
| Smoking | Current or former | 1.03 (0.70–1.52) | 1.41 (0.85–2.33) |
| Alcohol consumption (glasses/day) | Continuous | 0.90 (0.77–1.05) | 0.84 (0.70–1.01) |
| Coffee consumption (cups/day) | Continuous | 0.92 (0.82–1.02) | 0.90 (0.79–1.02) |
| >1 BCC at initial diagnosis | Yes | 1.49 (0.92–2.43) | 1.41 (0.79–2.52) |
| Localization of first BCC | Head and neck | Reference | Reference |
| Extremities | 18.07 (10.56–30.93)*** | 17.57 (10.06–30.70)*** | |
| Trunk | 11.44 (6.85–19.10)*** | 12.20 (7.08–21.03)*** |
*P‐value <0.05; ***P‐value <0.001.
†Compared to nodular, micronodular, infiltrative and mixed type BCCs; all mixed type BCCs with a superficial component were excluded. ‡Pooled ORs with 95% CIs between parentheses. §Full model, no selection procedures used.
BCC, basal cell carcinoma.
These associations remained strongly significant after the multivariable binary logistic regression analysis and no other predictors became significant (Table 2). Female gender gave an even stronger risk increase for sBCC (OR: 1.88, 95% CI: 1.16–3.03, P‐value <0.05), but localization remained the strongest predictor (truncal OR: 12.20, 95% CI: 7.08–21.03, P‐value <0.001; extremities OR: 17.57, 95% CI: 10.06–30.70, P‐value <0.001). The 11 predictors together explained 19.7% (Cox and Snell R2) of total variability of a superficial first BCC compared to a non‐superficial first BCC.
Study population for genetic analyses
Of the 14 628 Rotterdam Study participants linked to PALGA, 1257 were genotyped and had at least one BCC. After the exclusion of patients with a missing subtype (n = 181) and patients with a mixed superficial first BCC (n = 62), 1014 eligible BCC patients remained for further analyses. We randomly selected a BCC for participants with synchronous BCCs on their first diagnosis date (n = 126). Of the included patients, 159 (16%) had a superficial first BCC, 522 (51%) a nodular first BCC and the remaining 333 (33%) another subtype (infiltrative, micronodular or non‐superficial mixed type; Tables 3 and S3).
Table 3.
Genetic characteristics of 1014 Rotterdam Study patients with a first BCC
| Patient and tumour characteristics | Coding | Overalla | Superficial BCC | Non‐superficial BCC |
|---|---|---|---|---|
| Number of patients | 1014 (100%) | 159 (100%) | 855 (100%) | |
| Age at first BCC (years) | Median (IQR) | 72.9 (64.4–79.8) | 68.0 (60.8–75.6) | 73.5 (65.5–80.5) |
| Sex | Female | 556 (55%) | 103 (65%) | 453 (53%) |
Participants with a mixed type BCC with a superficial component were excluded.
BCC, basal cell carcinoma; IQR, interquartile range.
Patients with a superficial first BCC were younger than patients with a non‐superficial first BCC (median age 68.0 vs. 73.5 years), and the proportion females (65%) were higher in sBCC patients than in nsBCC patients (53%).
Genetic (SNP‐based) binary logistic regression analyses of sBCC vs. nsBCC
Of the 20 candidate SNPs, 2 were borderline significantly associated with a first sBCC in the univariable SNP‐based binary logistic regression analyses, namely rs8015138 (OR: 0.76, 95% CI: 0.60–0.97) and rs12203592 (OR: 1.55, 95% CI: 1.01–2.37; Table 4).
Table 4.
Associations between genetic predictors and occurrence of superficial first BCC (n = 1014)†
| Patient and tumour characteristics | Coding | Univariable models‡ | Multivariable models‡ , § | Multivariable model‡ , ¶ |
|---|---|---|---|---|
| Age at first BCC (years) | Continuous | 0.97 (0.95–0.98)*** | 0.96 (0.95–0.98)*** | |
| Sex | Female | 1.63 (1.15–2.32)** | 1.67 (1.16–2.40)** | |
| rs73635312 | Yes | 1.11 (0.74–1.67) | 1.08 (0.71–1.64) | 1.07 (0.70–1.63) |
| rs11170164 | Yes | 0.88 (0.57–1.35) | 0.93 (0.59–1.45) | 0.95 (0.61–1.49) |
| rs7335046 | Yes | 0.95 (0.67–1.36) | 0.92 (0.64–1.32) | 0.89 (0.61–1.29) |
| rs8015138 | Yes | 0.76 (0.60–0.97)* | 0.78 (0.61–1.00) | 0.79 (0.62–1.02) |
| rs1805007 | Yes | 0.96 (0.60–1.55) | 0.96 (0.59–1.57) | 0.95 (0.58–1.56) |
| rs78378222 | Yes | 0.89 (0.38–2.10) | 0.95 (0.40–2.52) | 0.98 (0.41–2.36) |
| rs7538876 | Yes | 1.06 (0.82–1.36) | 1.02 (0.79–1.32) | 1.02 (0.78–1.32) |
| rs801114 | Yes | 0.96 (0.75–1.23) | 1.03 (0.80–1.33) | 1.00 (0.77–1.30) |
| rs214782 | Yes | 0.89 (0.66–1.19) | 0.91 (0.68–1.22) | 0.93 (0.69–1.26) |
| rs13014235 | Yes | 1.07 (0.84–1.36) | 1.07 (0.83–1.37) | 1.07 (0.83–1.38) |
| rs57244888 | Yes | 1.06 (0.67–1.67) | 1.11 (0.70–1.78) | 1.12 (0.70–1.80) |
| rs401681 | Yes | 1.05 (0.82–1.33) | 1.00 (0.78–1.28) | 0.98 (0.76–1.26) |
| rs12203592 | Yes | 1.55 (1.01–2.37)* | 1.55 (1.01–2.39)* | 1.83 (1.13–2.97)* |
| rs12202284 | Yes | 0.92 (0.64–1.32) | 0.92 (0.63–1.33) | 0.72 (0.47–1.08) |
| rs12210050 | Yes | 1.00 (0.71–1.40) | 1.03 (0.73–1.46) | |
| rs157935 | Yes | 0.95 (0.73–1.24) | 0.94 (0.72–1.23) | 0.95 (0.72–1.26) |
| rs28727938 | Yes | 0.89 (0.52–1.53) | 0.78 (0.45–1.34) | 0.75 (0.43–1.32) |
| rs7006527 | Yes | 0.81 (0.58–1.13) | 0.79 (0.56–1.12) | 0.79 (0.55–1.12) |
| rs2151280 | Yes | 1.09 (0.86–1.38) | 1.13 (0.89–1.44) | 1.14 (0.89–1.46) |
| rs59586681 | Yes | 1.09 (0.85–1.40) | 1.13 (0.88–1.46) | 1.10 (0.85–1.44) |
*P‐value <0.05; **P‐value <0.01; ***P‐value <0.001.
†Compared to nodular, micronodular, infiltrative and mixed type BCCs; all mixed type BCCs with a superficial component were excluded.
‡Odds ratios with 95% confidence intervals between parentheses.
§Included one SNP at a time, adjusted for age at first BCC, sex and first 4 principal components.
¶Full model, adjusted for age at first BCC, sex and first 4 principal components. No selection procedures used.
BCC, basal cell carcinoma.
Before the multivariable SNP‐based binary logistic regression analyses, we excluded rs12210050 because it was highly correlated (Pearson's r: 0.86) with rs12202284 and both SNPs were also in strong linkage disequilibrium (r 2: 0.73) with each other. The multivariable analysis resulted in 1 promising SNP, namely rs12203592 (OR: 1.83, 95% CI: 1.13–2.97, P‐value 0.014) mapped to pigmentation gene IRF4, but after adjustment for multiple testing (FDR), this SNP lost its significance as well. No other SNPs were significantly associated with sBCC (Table 4).
The 19 candidate SNPs together explained 1.6%, of which rs12203592 explained 0.4% (Cox and Snell R2), of the total variability of a superficial first BCC compared to a non‐superficial first BCC.
Sensitivity analyses of sBCC vs. nodular BCC
After the non‐genetic multivariable binary logistic regression analysis comparing sBCC to nodular BCC, the same predictors (age at first BCC diagnosis, sex and a localization on the trunk or extremities) were significantly associated with a superficial first BCC with similar effect sizes (Table S4). The explained variability increased by 4.3% to 25.0% (Cox and Snell R2).
The multivariable SNP‐based binary logistic regression analysis comparing sBCC to nodular BCC resulted in 2 promising SNPs, namely rs12203592 (OR: 2.11, 95% CI: 1.25–3.58, P‐value 0.005) and rs12202284 (OR: 0.55, 95% CI: 0.35–0.88, P‐value 0.012), both mapped to the IRF4 – EXOC2 region, but were not in strong linkage disequilibrium (r 2: 0.18) with each other (Table S5). However, after adjustment for multiple testing (FDR) both SNPs lost their significance.
Discussion
This prospective population‐based cohort study replicates some previous non‐genetic findings and shows that there are significant differences between patients with a superficial first BCC and a non‐superficial first BCC. Patients who presented with a sBCC were younger, more often female and had their BCCs more frequently on the extremities and trunk than patients with nsBCCs. This study also looked into potential genetic differences. One SNP, mapped to IRF4, looked promising, but after adjustment for multiple testing, no significant differences in genetic make‐up between sBCC and nsBCC patients were found.
The associations found between the non‐genetic predictors and the occurrence of a superficial first BCC were in line with several other older and more recent observational studies from Europe and Australia.6, 7, 8, 9, 10, 13 However, most of these non‐genetic studies on histopathological BCC subtypes did not adjust for potential confounders.6, 7, 8, 9, 13 Therefore, it is possible that the associations found were spurious. We included 11 potential confounders in our non‐genetic multivariable model and found that patients with a superficial first BCC were significantly younger, almost twice as likely to be female and 12–18 times more likely to have their BCC on the trunk or extremities than patients with a non‐superficial first BCC. These differences in age, sex and localization could suggest that a different pattern of UVR exposure, namely intense intermittent, plays a role in the aetiology of sBCC as compared to nsBCC. A British and Australian cohort study showed that excessive recreational UVR exposure significantly increased the risk of truncal (superficial) BCCs,17, 29 whereas Dutch and Italian case–control studies showed no relation between cumulative lifetime UVR exposure and sBCC.10, 16
Another potential explanation for the significantly higher risk of sBCC in younger women could be behaviour. Women tend to use tanning beds more often than men30, 31 and pay closer attention to their health and physical appearance than men, which may lead to more medical visits.32
It is also possible that tumour biology differs at various anatomical sites. A Dutch renal transplant study showed that transplant recipients more often developed sBCCs and that their BCCs were located more frequently on the trunk and extremities than in the non‐immunosuppressed, which may point at role for the immune system.8
Superficial first BCC patients were significantly younger (approximately 5 years) than non‐superficial first BCC patients and developed their BCCs more often on relatively sun‐unexposed sites, which could mean that they have a different genetic predisposition which makes them more vulnerable to develop (superficial) BCC. It is possible that they, for example, have a reduced DNA repair capacity or other risk‐increasing DNA differences.33 Hence, we compared carefully selected BCC candidate SNPs between these two patient groups. Of the 19 included candidate SNPs in the multivariable regression analysis, rs12203592 looked most promising (OR: 1.83, 95% CI: 1.13–2.97, P‐value 0.014), but lost its significance after adjusting the FDR. This SNP is an intron variant mapped to the interferon regulatory factor 4 (IRF4) gene, which belongs to a well‐known family of transcription factors that are important in the regulation of the immune system. It is possible that certain SNPs downregulate the immune system which could lead to the formation of sBCC in relatively sun‐unexposed areas earlier in life. A recent genetic analysis of melanoma patients showed a significant association with the bimodal (early‐ and late‐onset) age distribution of melanoma for different rs12203592 genotypes.34 In addition, IRF4 also plays a key role in the pigmentation pathway and in the formation of (pre)malignancies of the skin.35, 36, 37 These premalignancies (i.e. actinic keratosis) have a superficial growth pattern which is comparable to that of sBCCs.
Limitations
Misclassification of BCC subtypes by pathologists most likely occurred throughout the study period, but it is unlikely that this misclassification was differential. However, we could not check the tissue samples as we only received excerpts from PALGA.
The total number of BCCs could have been underestimated, since we only included histopathologically confirmed BCCs. This underestimation will be most pronounced for superficial BCCs, because physicians could diagnose these BCCs visually and treat them non‐invasively. However, a recent Dutch observational study showed that only a small percentage (ca. 7%) of patients with metachronous BCCs had subsequent non‐histologically confirmed BCCs.38 In addition, the evidence‐based BCC guideline from the Dutch Society for Dermatology and Venereology states that histopathological verification is needed for all for BCC suspicious lesions.39 Finally, the distribution pattern of the histopathological subtypes in our study population is in line with other studies, with the nodular type being the most common, followed by the superficial type and infiltrative type, while mixed types were frequently found as well.6, 7, 8, 9, 10, 11 Our candidate SNP approach likely lacked sufficient power (26 degrees of freedom used and 159 patients with a superficial first BCC) despite the FDR approach taken. Detailed information about other limitations of the Rotterdam Study, the phenotype collection and the non‐genetic and genetic predictors can be found in two earlier publications.19, 40
Conclusion
Patients with a superficial first BCC differ from non‐superficial first BCC patients with respect to environmental factors (tumour localization as a proxy for UVR exposure) and phenotypic characteristics (age and sex), but (as far as we could find) not in genotype. As sBCC patients develop their first BCCs at a younger age, they could be at higher risk for subsequent skin cancers. Further study of the interplay between environmental, phenotypic and genotypic predictors and BCC subtypes may provide useful knowledge for BCC pathogenesis and the design of programs for prevention and early detection of BCC.
Supporting information
Table S1. Candidate single nucleotide polymorphisms for basal cell carcinoma or non‐melanoma skin cancer.
Table S2. Non‐genetic characteristics of 633 Rotterdam Study patients with a primary BCC.
Table S3. Genetic characteristics of 681 Rotterdam Study patients with a primary BCC.
Table S4. Associations between non‐genetic predictors and occurrence of superficial first BCC (n = 633).
Table S5. Associations between genetic predictors and occurrence of superficial first BCC (n = 681).
Acknowledgements
The Rotterdam Study is funded by Erasmus Medical Center and Erasmus University, Rotterdam, Netherlands Organization for the Health Research and Development (ZonMw), the Research Institute for Diseases in the Elderly (RIDE), the Ministry of Education, Culture and Science, the Ministry for Health, Welfare and Sports, the European Commission (DG XII), and the Municipality of Rotterdam. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. The authors are grateful to the study participants, the staff from the Rotterdam Study and the participating general practitioners and pharmacists. The generation and management of GWAS genotype data for the Rotterdam Study (RS I, RS II, RS III) were executed by the Human Genotyping Facility of the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC, Rotterdam, The Netherlands. The GWAS datasets are supported by the Netherlands Organisation of Scientific Research NWO Investments (nr. 175.010.2005.011, 911‐03‐012), the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC, the Research Institute for Diseases in the Elderly (014‐93‐015; RIDE2), the Netherlands Genomics Initiative (NGI)/Netherlands Organisation for Scientific Research (NWO) Netherlands Consortium for Healthy Aging (NCHA), project nr. 050‐060‐810. We thank Pascal Arp, Mila Jhamai, Marijn Verkerk, Lizbeth Herrera and Marjolein Peters, MSc, and Carolina Medina‐Gomez, MSc, for their help in creating the GWAS database, and Karol Estrada, PhD, Yurii Aulchenko, PhD, and Carolina Medina‐Gomez, MSc, for the creation and analysis of imputed data. We further thank Esther van den Broek and Lucy Overbeek from foundation PALGA, the Dutch Pathology Registry, for their help with the linkage. We also thank Senada Koljenovic for her help in the dermatopathology part of the linkage.
Conflicts of interest
The authors state no conflict of interest.
Funding source
This study was supported by a Netherlands Organization for the Health Research and Development (ZonMw) (no. 91711315).
References
- 1. Lomas A, Leonardi‐Bee J, Bath‐Hextall F. A systematic review of worldwide incidence of nonmelanoma skin cancer. Br J Dermatol 2012; 166: 1069–1080. [DOI] [PubMed] [Google Scholar]
- 2. Christenson LJ, Borrowman TA, Vachon CM et al Incidence of basal cell and squamous cell carcinomas in a population younger than 40 years. JAMA 2005; 294: 681–690. [DOI] [PubMed] [Google Scholar]
- 3. Flohil SC, van der Leest RJ, Arends LR et al Risk of subsequent cutaneous malignancy in patients with prior keratinocyte carcinoma: a systematic review and meta‐analysis. Eur J Cancer 2013; 49: 2365–2375. [DOI] [PubMed] [Google Scholar]
- 4. Hollestein LM, de Vries E, Aarts MJ et al Burden of disease caused by keratinocyte cancer has increased in The Netherlands since 1989. J Am Acad Dermatol 2014; 71: 896–903. [DOI] [PubMed] [Google Scholar]
- 5. Gordon LG, Rowell D. Health system costs of skin cancer and cost‐effectiveness of skin cancer prevention and screening: a systematic review. Eur J Cancer Prev 2015; 24: 141–149. [DOI] [PubMed] [Google Scholar]
- 6. Betti R, Inselvini E, Carducci M et al Age and site prevalence of histologic subtypes of basal cell carcinomas. Int J Dermatol 1995; 34: 174–176. [DOI] [PubMed] [Google Scholar]
- 7. McCormack CJ, Kelly JW, Dorevitch AP. Differences in age and body site distribution of the histological subtypes of basal cell carcinoma. A possible indicator of differing causes. Arch Dermatol 1997; 133: 593–596. [PubMed] [Google Scholar]
- 8. Bastiaens MT, Hoefnagel JJ, Bruijn JA et al Differences in age, site distribution, and sex between nodular and superficial basal cell carcinoma indicate different types of tumors. J Invest Dermatol 1998; 110: 880–884. [DOI] [PubMed] [Google Scholar]
- 9. Scrivener Y, Grosshans E, Cribier B. Variations of basal cell carcinomas according to gender, age, location and histopathological subtype. Br J Dermatol 2002; 147: 41–47. [DOI] [PubMed] [Google Scholar]
- 10. Pelucchi C, Di Landro A, Naldi L et al Risk factors for histological types and anatomic sites of cutaneous basal‐cell carcinoma: an italian case‐control study. J Invest Dermatol 2007; 127: 935–944. [DOI] [PubMed] [Google Scholar]
- 11. Betti R, Radaelli G, Crosti C et al Margin involvement and clinical pattern of basal cell carcinoma with mixed histology. J Eur Acad Dermatol Venereol 2012; 26: 483–487. [DOI] [PubMed] [Google Scholar]
- 12. Rippey JJ. Why classify basal cell carcinomas? Histopathology 1998; 32: 393–398. [DOI] [PubMed] [Google Scholar]
- 13. Arits AH, Schlangen MH, Nelemans PJ et al Trends in the incidence of basal cell carcinoma by histopathological subtype. J Eur Acad Dermatol Venereol 2011; 25: 565–569. [DOI] [PubMed] [Google Scholar]
- 14. Kopf AW. Computer analysis of 3531 basal‐cell carcinomas of the skin. J Dermatol 1979; 6: 267–281. [DOI] [PubMed] [Google Scholar]
- 15. Lovatt TJ, Lear JT, Bastrilles J et al Associations between ultraviolet radiation, basal cell carcinoma site and histology, host characteristics, and rate of development of further tumors. J Am Acad Dermatol 2005; 52: 468–473. [DOI] [PubMed] [Google Scholar]
- 16. Kennedy C, Bajdik CD, Willemze R et al The influence of painful sunburns and lifetime sun exposure on the risk of actinic keratoses, seborrheic warts, melanocytic nevi, atypical nevi, and skin cancer. J Invest Dermatol 2003; 120: 1087–1093. [DOI] [PubMed] [Google Scholar]
- 17. Lovatt TJ, Lear JT, Bastrilles J et al Associations between UVR exposure and basal cell carcinoma site and histology. Cancer Lett 2004; 216: 191–197. [DOI] [PubMed] [Google Scholar]
- 18. Hofman A, Brusselle GG, Darwish Murad S et al The Rotterdam Study: 2016 objectives and design update. Eur J Epidemiol 2015; 30: 661–708. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Verkouteren JA, Smedinga H, Steyerberg EW et al Predicting the risk of a second basal cell carcinoma. J Invest Dermatol 2015; 135: 2649–2656. [DOI] [PubMed] [Google Scholar]
- 20. Casparie M, Tiebosch AT, Burger G et al Pathology databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide histopathology and cytopathology data network and archive. Cell Oncol 2007; 29: 19–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. LeBoit PE, Burg G, Weedon D et al World Health Organization Classification of Tumours. Pathology and Genetics of Skin Tumours. IARC Press, Lyon, 2006. [Google Scholar]
- 22. De Hertog SA, Wensveen CA, Bastiaens MT et al Relation between smoking and skin cancer. J Clin Oncol 2001; 19: 231–238. [DOI] [PubMed] [Google Scholar]
- 23. Husein‐Elahmed H, Aneiros‐Fernandez J, Gutierrez‐Salmeron MT et al Alcohol intake and risk of aggressive histological basal cell carcinoma: a case‐control study. Eur J Dermatol 2012; 22: 525–530. [DOI] [PubMed] [Google Scholar]
- 24. Song F, Qureshi AA, Han J. Increased caffeine intake is associated with reduced risk of basal cell carcinoma of the skin. Cancer Res 2012; 72: 3282–3289. [DOI] [PubMed] [Google Scholar]
- 25.In Vol. 2015: National Human Genome Research Institute (NHGRI) and the European Bioinformatics Institute (EMBL‐EBI). URL http://www.ebi.ac.uk/gwas/ (last accessed: 16 December 2015).
- 26. Yang J, Lee SH, Goddard ME et al GCTA: a tool for genome‐wide complex trait analysis. Am J Hum Genet 2011; 88: 76–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Hoaglin DC, Iglewicz B. Fine‐tuning some resistant rules for outlier labeling. J Am Stat Assoc 1987; 82: 1147–1149. [Google Scholar]
- 28. Benjamini Y, Hochberg Y. Controlling the false discovery rate ‐ a practical and powerful approach to multiple testing. J R Stat Soc B Methodol 1995; 57: 289–300. [Google Scholar]
- 29. Neale RE, Davis M, Pandeya N et al Basal cell carcinoma on the trunk is associated with excessive sun exposure. J Am Acad Dermatol 2007; 56: 380–386. [DOI] [PubMed] [Google Scholar]
- 30. Robinson JK, Rigel DS, Amonette RA. Trends in sun exposure knowledge, attitudes, and behaviors: 1986 to 1996. J Am Acad Dermatol 1997; 37: 179–186. [DOI] [PubMed] [Google Scholar]
- 31. Koster B, Thorgaard C, Clemmensen IH et al Sunbed use in the Danish population in 2007: a cross‐sectional study. Prev Med 2009; 48: 288–290. [DOI] [PubMed] [Google Scholar]
- 32. Swetter SM, Layton CJ, Johnson TM et al Gender differences in melanoma awareness and detection practices between middle‐aged and older men with melanoma and their female spouses. Arch Dermatol 2009; 145: 488–490. [DOI] [PubMed] [Google Scholar]
- 33. Wei Q, Matanoski GM, Farmer ER et al DNA repair and aging in basal cell carcinoma: a molecular epidemiology study. Proc Natl Acad Sci USA 1993; 90: 1614–1618. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Gibbs DC, Orlow I, Bramson JI et al Association of interferon regulatory factor‐4 polymorphism rs12203592 with divergent melanoma pathways. J Natl Cancer Inst 2016; 108: djw004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Han J, Kraft P, Nan H et al A genome‐wide association study identifies novel alleles associated with hair color and skin pigmentation. PLoS Genet 2008; 4: e1000074. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Gerstenblith MR, Shi J, Landi MT. Genome‐wide association studies of pigmentation and skin cancer: a review and meta‐analysis. Pigment Cell Melanoma Res 2010; 23: 587–606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Jacobs LC, Liu F, Pardo LM et al IRF4, MC1R and TYR genes are risk factors for actinic keratosis independent of skin color. Hum Mol Genet 2015; 24: 3296–3303. [DOI] [PubMed] [Google Scholar]
- 38. Flohil SC, van Tiel S, Koljenovic S et al Frequency of non‐histologically diagnosed basal cell carcinomas in daily Dutch practice. J Eur Acad Dermatol Venereol 2013; 27: 907–911. [DOI] [PubMed] [Google Scholar]
- 39. Dutch Society for Dermatology and Venereology (NVDV) , Evidence‐based guideline basal cell carcinoma. Utrecht, the Netherlands, 2015. [Google Scholar]
- 40. Verkouteren JA, Pardo LM, Uitterlinden AG et al Common variants affecting susceptibility to develop multiple basal cell carcinomas. J Invest Dermatol 2015; 135: 2135–2138. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. Candidate single nucleotide polymorphisms for basal cell carcinoma or non‐melanoma skin cancer.
Table S2. Non‐genetic characteristics of 633 Rotterdam Study patients with a primary BCC.
Table S3. Genetic characteristics of 681 Rotterdam Study patients with a primary BCC.
Table S4. Associations between non‐genetic predictors and occurrence of superficial first BCC (n = 633).
Table S5. Associations between genetic predictors and occurrence of superficial first BCC (n = 681).