Non-genetic risk factors for cutaneous melanoma and keratinocyte skin cancers: an

umbrella review of meta-analyses

Lazaros Belbasis1, Irene Stefanaki2, Alexander J. Stratigos2, Evangelos Evangelou1,3

1Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina,

Greece

2Department of Dermatology, Andreas Sygros Hospital, University of Athens Medical School,

Athens, Greece

3Department of Biostatistics and Epidemiology, School of Public Health, Imperial College

London, London, UK

Word count: 3,498

Funding sources: None

Number of references: 78

Number of figures: 1

Number of tables: 2

Number of supplementary tables: 1

1

Corresponding author:

Dr Evangelos Evangelou

Assistant Professor

Department of Hygiene and Epidemiology,

University of Ioannina Medical School,

University Campus, Ioannina, Greece

Tel: +302651007720

e-mail: vangelis@cc.uoi.gr

2

Abstract

Background: Skin cancers have a complex disease mechanism, involving both genetic and non-

genetic risk factors. Numerous meta-analyses have been published claiming statistically

significant associations between non-genetic risk factors and skin cancers without applying a

thorough methodological assessment.

Objective: The present study maps the literature on the non-genetic risk factors of skin cancers,

assesses the presence of statistical biases and identifies the associations with robust evidence.

Methods: We searched PubMed up to January 20, 2016 to identify systematic reviews and meta-

analyses of observational studies that examined associations between non-genetic factors and

skin cancers. For each meta-analysis, we estimated the summary effect size by random-effects

and fixed-effects models, the 95% confidence interval and the 95% prediction interval. We also

assessed the between-study heterogeneity (I2 metric), evidence for small-study effects and excess

significance bias.

Results: Forty-four eligible papers were identified and included a total of 85 associations.

Twenty-one associations were significant at P<10-6. Fifty-two associations had large or very

large heterogeneity. Evidence for small-study effects and excess significance bias was found in

fifteen and thirteen associations, respectively. Overall, thirteen associations (actinic keratosis,

serum vitamin D, sunburns, and hair color for basal cell carcinoma and density of freckles, eye

color, hair color, history of melanoma, skin type, sunburns, premalignant skin lesions, common

and atypical nevi for melanoma) presented high level of credibility.

Conclusion: The majority of meta-analyses on non-genetic risk factors for skin cancers suffered

from large between-study heterogeneity and small-study effects or excess significance bias. The

3

associations with convincing and highly suggestive evidence were mainly focused on skin

photosensitivity and phenotypic characteristics.

Keywords: basal cell carcinoma, risk factors, melanoma, keratinocyte skin cancers, skin cancers,

squamous cell carcinoma

Abbreviations: AIDS: acquired immune deficiency syndrome, BCC: basal cell carcinoma, CI:

confidence interval, CM: cutaneous melanoma, HPV: human papilloma virus, IQR: interquartile

range, KSCs: keratinocyte skin cancers, NSAIDs: non-steroid anti-inflammatory drugs, SCC:

squamous cell carcinoma, SE: standard error, UV: ultraviolet

4

Introduction

Skin cancers are grouped into two discrete categories: cutaneous melanoma (CM) and

keratinocyte skin cancers (KSCs). CM is a malignancy that derives from melanocytes, i.e.

pigment-producing cells of neuroectodermal origin that can be found primarily in the skin. [1]

The incidence of CM presents a geographic variation with the highest incidence rates reported in

Australia and New Zealand, followed by the USA and certain areas in Europe. [2] The term

“keratinocyte skin cancers” involves a wide cluster of skin cancers excluding melanoma,

although it is mainly used to describe epithelial skin cancers, e.g. basal cell carcinomas (BCC)

and squamous cell carcinomas (SCC). [3] The epidemiology of KSCs presents a similar

geographic variation and the higher incidence is observed in countries with high average altitude

and high levels of ultraviolet radiation. [4] KSCs are the most common human cancers, and their

incidence continues to rise in a global level. [5] Both CM and KSCs have a complex etiology

involving environmental, phenotypical and genetic risk factors.

The genetic predisposition of CM and KSCs has already been investigated by candidate-

gene association studies, genome-wide association studies and field synopses. [6–12] With

respect to extrinsic factors, exposure to ultraviolet (UV) light is considered the main causative

risk factor for both types of malignant skin tumors. Certain phenotypic characteristics have also

been associated with an increased risk for CM. Also, some additional factors, such as artificial

UV sources, and lifestyle factors have been linked with an increased risk of skin cancers in the

literature, without always providing conclusive evidence. [13]

Numerous meta-analyses for phenotypic and environmental factors associated with CM

and KSCs have been published without, however, having been critically appraised. In the present

5

study, we aim to map the literature on the non-genetic risk factors that have been associated with

skin cancers and to evaluate whether there is evidence for diverse biases in this literature. We

additionally highlight which of the associations that have already been considered in meta-

analyses present strong evidence for association.

6

Methods

Search strategy and eligibility criteria

We conducted an umbrella review, which is a systematic collection and assessment of

multiple systematic reviews and meta-analyses performed on a specific research topic. [14] The

methods of the umbrella review are standardized and constitute a state-of-the-art approach which

has already been applied in the assessment of environmental risk factors in the field of

neurodegenerative diseases. [15–17]

We performed a systematic review on PubMed from inception to January 20, 2016 to

identify systematic reviews and meta-analyses of observational studies examining associations

between environmental factors and both CM and KSCs. The search strategy included the

following keywords: (melanoma OR “skin cancer” OR “squamous cell carcinoma” OR “basal

cell carcinoma”) AND (“systematic review” OR meta-analysis OR pooled). We excluded meta-

analyses that investigated the association between genetic markers and risk of skin cancers,

because these factors are already summarized, evaluated and publicly available in a regularly

updated field synopsis of genetic association studies and genome-wide association studies. [6–

12] Also, we excluded meta-analyses that had less than 3 observational studies; that did not

provide data on effect estimates of individual component studies; that were synthesized

prevention randomized clinical trials; or that examined prognostic factors of skin cancers. We

did not apply any language restrictions in the selection of eligible studies. When more than one

meta-analysis on the same research question was available, we kept the meta-analysis with the

largest number of component studies and the largest total number of cases.

Data extraction

7

From each eligible article, we extracted the following information: first author, journal,

year of publication, examined risk factors, and number of studies included. We also recorded the

number of cases and controls in each study for each risk factor, and the study-specific relative

risk estimates (risk ratio, odds ratio, standardized incidence ratio, or hazard ratio) or standardized

mean differences along with the corresponding confidence intervals (CI). When the sample sizes

of the component studies were not available through the article of meta-analysis, we retrieved the

published report of the component study and we extracted the relevant data.

Statistical analysis

For each meta-analysis, we used both fixed-effects and random-effects models to

estimate the summary effect size and its 95% CI. [18,19] We estimated the 95% prediction

interval, which further accounts for between-study heterogeneity and evaluates the uncertainty

for the effect that would be expected in a new study addressing that same association. [20,21]

For the largest study of each meta-analysis, we calculated the standard error (SE) of the effect

size, we examined whether the standard error was less than 0.10 and whether the largest study

presented a statistically significant effect. In a study with SE of less than 0.10, the difference

between the effect estimate and the upper or lower 95% confidence interval is less than 0.20 (i.e.

this uncertainty is less than what is considered a small effect size).

We estimated the I2 metric to quantify the between-study heterogeneity. I2 ranges

between 0% and 100% and it is the ratio of between-study variance over the sum of the within-

study and between-study variances. [22] Values exceeding 50% or 75% are considered to

represent large or very large heterogeneity, respectively. [23] Also, we used the Kruskal-Wallis

test to examine whether there is a statistically significant difference in median I 2 values per

8

phenotype (i.e., basal cell carcinoma, cutaneous melanoma, squamous cell carcinoma, and both

types of keratinocyte skin cancers). We claimed a statistically significant difference at P<0.05.

The presence of small-study effects was assessed by the regression asymmetry test

proposed by Egger and colleagues. [24,25] We claimed that small-study effects were present in a

meta-analysis, when Egger test was statistically significant at P<0.10 with a more conservative

effect in the largest study than in random-effects meta-analysis.

The excess statistical significance test was performed to assess whether the observed

number of studies with nominally significant results is larger than their expected number. [26]

The expected number of studies with significant results is calculated in each meta-analysis by

summing the statistical power estimates for each component study. As previously proposed, the

power of each component study was estimated using the effect size of the largest study (smallest

SE) in a meta-analysis. [27] The power of each study was calculated with an algorithm using a

non-central t distribution. [28] Excess statistical significance for single meta-analyses was

claimed at P<0.10. [26]

Power calculations and excess statistical significance tests were not performed for the

meta-analyses including record linkage studies, because these studies often did not adequately

report the sample sizes. These meta-analyses pertained to the following 15 risk factors: systemic

lupus erythematosus, [29] organ transplantation, [30] occupational exposure to UV light, [31,32]

chronic sun exposure, [33] history of melanoma, [34] non Hodgkin lymphoma, [35] chronic

lymphocytic leukemia, [36] Parkinson’s disease, [37] AIDS, [38,39] psoriasis, [40] airline pilots

and cabin crew, [41] Merkel cell carcinoma, [42] rheumatoid arthritis, [43] IBD [44] and TNF

inhibitors. [45]

9

Assessment of epidemiological credibility

We graded the statistically significant associations according to predefined criteria to

identify the associations that had the strongest validity and were not suggestive of bias.

Specifically, we graded as convincing the associations fulfilling the following criteria: had a

significant effect under the random-effects model at P<10-6, [46] were based on more than 1,000

cases, had not large between-study heterogeneity (I2<50%), had 95% prediction interval that

excluded the null value, and absence of evidence for small-study effects or excess significance

bias. We considered as highly suggestive the associations with more than 1,000 cases, a

significant effect at P<10-6, and a statistically significant effect present at the largest study. The

associations with a significant effect at P<10-3 and more than 1,000 cases were considered as

suggestive. The rest of the significant associations (P<0.05) were classified as having weak

evidence.

The statistical analyses and the power calculations were done with STATA version 12.0.

10

Results

Overall, 1,922 articles were identified by the search strategy, and 44 articles were deemed

eligible (Figure 1). Forty of the articles screened by full-text were excluded because another

meta-analysis with a larger number of component studies was available. The publication date of

the eligible articles ranged between 2005 and 2016. The 44 eligible articles included 85

associations (16 for BCC, 53 for cutaneous melanoma, 9 for SCC and 7 for both types of KSCs).

Four articles focused exclusively on prospective cohort studies, [47–50] 16 articles included

record-linkage studies, [29–45] whereas the rest articles included a mix of cross-sectional, case-

control, and cohort studies.

Summary effects and significant findings

Under the random-effects model, 58 associations (68%) presented a significant effect at

P<0.05 (Table 1). Thirty-five associations (41%) presented a P<0.001, while only 21 (25%)

survived after the application of a more stringent p-value threshold (P<10-6). Five risk factors

(serum vitamin D, hair color, eye color, sunburns, and actinic keratosis) presented a P<10 -6 for an

association with BCC (Table 1). Fifteen risk factors with a P<10-6 were associated with CM and

pertained to: density of freckles, eye color, hair color, premalignant skin lesions, skin color, skin

type, history of sunburns, common nevi, organ transplantation, history of melanoma, atypical

nevi, airline pilots and cabin crew, Merkel cell carcinoma, Parkinson’s disease and non-Hodgkin

lymphoma (Table 1). Only one risk factor (occupational UV light exposure) presented a

significant effect at 10-6 for an association with SCC (Table 1). The median number of datasets

per meta-analysis was 10 (IQR, 7-18) and the median number of cases was 4,000 (IQR, 1,601-

11

7,804). The number of cases was greater than 1,000 in 63 meta-analyses. In 51 associations

(60%) the SE was less than 0.10. (Supplementary Table 1).

Between-study heterogeneity and prediction intervals

Thirty-three (39%) of the examined associations presented not large heterogeneity

(I2<50%), twenty-four associations (28%) had large heterogeneity (I2≥50% and I2≤75%), and the

rest 28 associations (33%) had very large heterogeneity (I2>75%). In seventeen associations

(20%), the 95% prediction interval under the random-effects model did not include the null value

(Table 1). These associations pertained to serum vitamin D, hair color, sunburns, and actinic

keratosis for risk of BCC; birth weight, density of freckles, hair color, premalignant skin lesions,

organ transplantation, history of melanoma, non-Hodgkin lymphoma, atypical nevi, common

nevi, airline pilots and cabin crews, and retinol intake for risk of CM; TNF inhibitors for

keratinocyte skin cancers; and non-aspirin NSAIDs for SCC. No statistically significant

difference was observed in the median I 2 values across the different phenotypes (P=0.473).

Small-study effects and excess significance bias

Both criteria for the presence of small-study effects (statistically significant Egger’s test

at P<0.10 and largest study with a significant effect) were fulfilled in 15 associations (Table 1,

Supplementary Table 1). These associations pertained to occupational ultraviolet light exposure,

eye color, skin color, freckles in childhood, solar lentigines, and aspirin for BCC; occupational

ultraviolet light exposure for SCC; and serum vitamin D, eye color, intermittent sun exposure,

skin color, skin type, sunburns, Parkinson’s disease, and smoking for melanoma. Thirteen

associations had hints for excess significance bias (aspirin or non-aspirin NSAIDs, eye color,

skin color, freckles in childhood, and solar lentigines for BCC; intermittent sun exposure, skin

12

color, sunburns, common nevi, smoking, sunscreen use for CM; smoking and β-genus HPV

infection for SCC) (Supplementary Table 1).

Epidemiological credibility of findings

Four associations had a significant effect at P<10-6, more than 1,000 cases, not large

heterogeneity and 95% prediction interval excluding the null value, and no hints for small-study

effects and excess significance bias. The associations with convincing evidence are serum

vitamin D, hair color, and actinic keratosis for BCC, and hair color for CM. Eight risk factors

(density of freckles, eye color, premalignant skin lesions, skin type, sunburns, common nevi,

atypical nevi and history of melanoma) for CM and one association (sunburns) for BCC

presented highly suggestive evidence. Also, 11 risk factors were supported by suggestive

evidence, and thirty-four associations presented weak evidence (Table 2).

13

Discussion

In this large umbrella review, we systematically mapped and appraised all the evidence,

derived from meta-analyses, for non-genetic associations with CM and KSCs, using state-of-the

art evidence synthesis and bias appraisal approaches. [6,7,15–17,51] Overall, we examined 16

associations for BCC, 53 associations for CM, 9 for SCC and 7 for both types of KSCs. The

majority of meta-analyses on non-genetic risk factors for skin cancers suffered from large

between-study heterogeneity and small-study effects or excess significance bias, indicating the

real challenges of meta-analyses of observational studies to evaluate and establish true and

precise risk estimates. Of course, large heterogeneity does not necessarily indicate a false

positive finding but may definitely affect the precision of the risk estimate.

The risk factors that achieved convincing or highly suggestive evidence for an association

with CM were mainly related to sun exposure and skin photosensitivity, supporting the long-

standing notion that excess UV exposure is the main extrinsic factor for developing melanoma.

In our analysis, four phenotypic characteristics (hair color, density of freckles, eye color, and

skin type) demonstrated convincing or highly suggestive evidence for an increased risk of

melanoma. These factors are strongly related to the amount and type of cutaneous melanin and

the sensitivity to UV light. [52] Skin color is considered a principal factor responsible for

melanoma, while hair and eye color are proxy of skin phenotype, since all of them are dependent

on the amount of cutaneous melanin. [52] This is supported by findings on genetic risk factors

where several genetic variants associated with an increased risk for melanoma were also linked

with specific pigmentation characteristics. [6,7,13]

14

Furthermore, common and atypical nevi were supported by highly suggestive evidence

for an association with melanoma. A dose-response meta-analysis taking into account the count

of nevi showed that the number of nevi is correlated to the magnitude of risk for CM; an

increased count of either common or atypical nevi was associated with a higher risk for CM. [53]

The true effect of these associations could be inflated, because epidemiological studies on nevi

use different and non-standardized approaches to measure the number of nevi and these

techniques are considered quite imprecise. [54]

The different aspects of sun exposure (total, intermittent or chronic sun exposure) and

their association to CM susceptibility did not present definitive convincing or highly suggestive

evidence in our umbrella review. The meta-analyses examining these associations had large or

very large heterogeneity and 95% prediction interval that included the null value. This could be

mainly attributed to the large variability in recording and coding this variable in the

observational studies, reflecting the challenges in the design and analyses of such studies. [33]

Additionally, the history of sunburns presented highly suggestive evidence of association with

risk for CM. However, the majority of available evidence for the association of sun exposure and

sunburns with CM was retrospective case-control studies, a study design prone to systematic

biases, such as recall bias. Indeed, patients with CM are more likely to retrospectively self-report

an excess history of multiple sunburns and high level of sun exposure due to heightened public

awareness compared to healthy controls. Also, airline pilots and cabin crew presented an

increased risk for CM, and this association could be a proxy of higher exposure to UV light of

this occupational group compared to the general population. This association was supported by

weak evidence due to the presence of large between-study statistical heterogeneity and a trivial

number of cases.

15

Likewise, premalignant skin lesions and history of prior melanoma presented highly

suggestive evidence and their effect was impressively large. Premalignant skin lesions, a wide

cluster of conditions including actinic keratosis, were associated with risk for CM, presenting a

large risk ratio and very low p-value, but this association had very large heterogeneity. The

history of prior melanoma presented a standardized incidence ratio larger than 10. However, the

between-study heterogeneity, as quantified by I2, was more than 90%; thus, this association

should be interpreted with caution and potential sources of heterogeneity should be explored.

The true effect of history of prior melanoma might be lower, but it still represents a major risk

factor for developing a subsequent CM. This is an observation of great importance for public

health interventions. [34] People with a history of prior melanoma should be aware of this fact

and their general practitioners should carefully examine the patients for early identification of a

new melanoma.

The association between elevated serum levels of 25(OH)D (a biomarker of vitamin D in

serum) and risk of BCC presented convincing evidence. This association probably is not due to a

harmful effect of vitamin D. Indeed, the serum levels of 25(OH)D are correlated with the level of

sun exposure and the sun exposure is considered a causal factor for BCC, although this

association has not been assessed through a meta-analytic synthesis. Also, our analysis indicated

that there is convincing evidence for an association of hair color, and actinic keratosis with risk

of BCC. As already mentioned above, hair color constitutes a proxy of skin photosensitivity,

while actinic keratosis is a premalignant skin lesion induced by chronic exposure to UV light.

[55,56] The potential role of sun exposure in the risk for BCC is further supported by the

association between sunburns and risk for BCC, an association with highly suggestive evidence.

As in the case of CM, the other skin phenotypic characteristics (eye and skin color) were

16

characterized by large or very large heterogeneity and/or presence of small-study effects and

excess significance bias.

Several studies have investigated the association of indoor tanning with CM and KSCs

showing an overall weak effect. Indoor tanning has been classified by the International Agency

for Research on Cancer among the highest category of carcinogens, and a recommendation

discouraging persons younger than 30 years old to use sunbeds has been published. [57] This

decision was based on the findings of a meta-analysis which indicated that the overall effect of

indoor tanning on the risk for developing melanoma was non-significant, but the effect became

significant with an about 2-fold increase in the risk of CM, when the first exposure to indoor

tanning occurred during adolescence and early adult life. [58] In our assessment of the

association between indoor tanning and CM, we focused on a more recent meta-analysis

including more observational studies. [59] In this meta-analysis, the association presented a

statistically significant effect. However, the association was supported by weak evidence,

because it had large heterogeneity with a 95% prediction interval including the null value, and

the largest study presented a non-significant effect. The body of evidence on indoor tanning and

risk for CM was mainly consisted of case-control studies with low to moderate quality.

Similarly, the association between indoor tanning and BCC was supported by weak evidence.

However, highly suggestive evidence exists for an increased risk of SCC in individuals exposed

to indoor tanning. Although the available epidemiological evidence did not lead to a thorough

conclusion, indoor tanning is a recreational activity. The presence of biases and the retrospective

design of studies could inflate the effect size, but a genuine association could not be excluded.

Thus, the recommendation of international scientific societies should not be doubted as a public

health intervention.

17

Our study was able to examine other controversial associations to the extent that these

have been considered by published meta-analyses. Specifically, chronic immunosuppressant

conditions (organ transplant recipients and AIDS) and chronic inflammatory diseases associated

with immunosuppressant medication (psoriasis, rheumatoid arthritis and inflammatory bowel

disease) showed a significant association with melanoma but the evidence was weak (very small

number of cases, and large or very large between-study heterogeneity). Additionally, patients

with Parkinson’s disease had an increased risk for cutaneous melanoma, but this association was

supported by less than 1,000 cases of melanoma, very large heterogeneity and presence of small-

study effects.

Overall, our umbrella review indicated that more than half of the associations for non-

genetic risk factors of skin cancers had evidence for large or very large between-study

heterogeneity. Furthermore, only one fifth of the associations had a 95% prediction interval

excluding the null value. In many cases, the large heterogeneity estimates could be explained by

the absence of standardized methods to accurately measure and ascertain the exposures, and the

between-studies variability in the clinical measurements and methods applied for the calculation

of the associations. The presence of heterogeneity does not necessarily imply that the observed

associations are false positive signals, but it highlights the need for standardized and accurate

methods to measure the traditional risk factors of skin cancers (i.e., sun exposure, cutaneous nevi

and phenotypic characteristics).

Our analysis has some limitations. First, we did not appraise the quality of the component

studies, because this was beyond the scope of this umbrella review. This should be the aim of the

original systematic reviews and meta-analyses, which should include an assessment of study

quality and examine whether the component studies should be included in the quantitative

18

calculations, by assessing the methodological quality of the studies. Also, in our analysis we

considered only associations examined by meta-analyses of observational studies. Thus, we

might miss other associations supported by adequate evidence that have not yet been assessed

through meta-analytic approaches. Furthermore, some associations were studied in pooled

analyses of individual participant data and study-specific effect estimates were not publicly

available. Additionally, we could not perform the excess significance test for some associations

due to under-reporting of sample sizes in original studies. However, given that these associations

were mainly supported by a small number of cases or presented large between-study

heterogeneity, this fact did not affect our appraisal of epidemiological credibility.

In this study, we provided a mapping of non-genetic risk factors examined in meta-

analyses for CM and KSCs, showing that phenotypic characteristics associated with sensitivity to

UV light exposure have highly suggestive or convincing evidence for an association with CM

risk. Also, most of these associations have an apparent biological plausibility. Furthermore, high

levels of serum 25(OH)D were associated with an increased risk of BCC, although this

association probably could be profoundly attributed to the sun exposure. Also, red hair and

presence of actinic keratosis had convincing evidence for an increased risk of BCC. Interestingly

no association was confirmed for cutaneous SCC, presumably due to the small number of meta-

analyses focusing on this tumor. Our study indicates that the observational research in the field

of skin cancers suffers from the presence of large between-study heterogeneity and statistical

biases that hinder the identification of robust risk factors. Standardization and harmonization of

datasets will allow for the computation of more precise estimates and will promote the

development and training of prediction models that could promote public health.

19

References

[1] D. Schadendorf, D.E. Fisher, C. Garbe, J.E. Gershenwald, J.-J. Grob, A. Halpern, et al., Melanoma, Nat. Rev. Dis. Prim. 1 (2015) 15003. doi:10.1038/nrdp.2015.3.

[2] F. Erdmann, J. Lortet-Tieulent, J. Schüz, H. Zeeb, R. Greinert, E.W. Breitbart, et al., International trends in the incidence of malignant melanoma 1953-2008--are recent generations at higher or lower risk?, Int. J. Cancer. 132 (2013) 385–400. doi:10.1002/ijc.27616.

[3] L.A. Kwasniak, J. Garcia-Zuazaga, Basal cell carcinoma: evidence-based medicine and review of treatment modalities, Int. J. Dermatol. 50 (2011) 645–658. doi:10.1111/j.1365-4632.2010.04826.x.

[4] A. Lomas, J. Leonardi-Bee, F. Bath-Hextall, A systematic review of worldwide incidence of nonmelanoma skin cancer., Br. J. Dermatol. 166 (2012) 1069–80. doi:10.1111/j.1365-2133.2012.10830.x.

[5] V. Madan, J.T. Lear, R.-M. Szeimies, Non-melanoma skin cancer, Lancet. 375 (2010) 673–685. doi:10.1016/S0140-6736(09)61196-X.

[6] F. Chatzinasiou, C.M. Lill, K. Kypreou, I. Stefanaki, V. Nicolaou, G. Spyrou, et al., Comprehensive field synopsis and systematic meta-analyses of genetic association studies in cutaneous melanoma., J. Natl. Cancer Inst. 103 (2011) 1227–35. doi:10.1093/jnci/djr219.

[7] K. Antonopoulou, I. Stefanaki, C.M. Lill, F. Chatzinasiou, K.P. Kypreou, F. Karagianni, et al., Updated field synopsis and systematic meta-analyses of genetic association studies in cutaneous melanoma: the MelGene database., J. Invest. Dermatol. 135 (2015) 1074–9. doi:10.1038/jid.2014.491.

[8] T. Rafnar, P. Sulem, S.N. Stacey, F. Geller, J. Gudmundsson, A. Sigurdsson, et al., Sequence variants at the TERT-CLPTM1L locus associate with many cancer types., Nat. Genet. 41 (2009) 221–7. doi:10.1038/ng.296.

[9] S.N. Stacey, P. Sulem, D.F. Gudbjartsson, A. Jonasdottir, G. Thorleifsson, S.A. Gudjonsson, et al., Germline sequence variants in TGM3 and RGS22 confer risk of basal cell carcinoma., Hum. Mol. Genet. 23 (2014) 3045–53. doi:10.1093/hmg/ddt671.

[10] H. Nan, M. Xu, P. Kraft, A.A. Qureshi, C. Chen, Q. Guo, et al., Genome-wide association study identifies novel alleles associated with risk of cutaneous basal cell carcinoma and squamous cell carcinoma., Hum. Mol. Genet. 20 (2011) 3718–24. doi:10.1093/hmg/ddr287.

[11] M. Zhang, F. Song, L. Liang, H. Nan, J. Zhang, H. Liu, et al., Genome-wide association studies identify several new loci associated with pigmentation traits and skin cancer risk in European Americans., Hum. Mol. Genet. 22 (2013) 2948–59. doi:10.1093/hmg/ddt142.

[12] M.H. Law, D.T. Bishop, J.E. Lee, M. Brossard, N.G. Martin, E.K. Moses, et al., Genome-wide meta-analysis identifies five new susceptibility loci for cutaneous malignant melanoma., Nat. Genet. 47 (2015) 987–95. doi:10.1038/ng.3373.

20

[13] A.M. Eggermont, A. Spatz, C. Robert, Cutaneous melanoma, Lancet. 383 (2014) 816–827. doi:10.1016/S0140-6736(13)60802-8.

[14] J.P.A. Ioannidis, Integration of evidence from multiple meta-analyses: a primer on umbrella reviews, treatment networks and multiple treatments meta-analyses., CMAJ. 181 (2009) 488–93. doi:10.1503/cmaj.081086.

[15] L. Belbasis, V. Bellou, E. Evangelou, J.P.A. Ioannidis, I. Tzoulaki, Environmental risk factors and multiple sclerosis: an umbrella review of systematic reviews and meta-analyses, Lancet Neurol. 14 (2015) 263–273. doi:10.1016/S1474-4422(14)70267-4.

[16] V. Bellou, L. Belbasis, I. Tzoulaki, E. Evangelou, J.P.A. Ioannidis, Environmental risk factors and Parkinson’s disease: An umbrella review of meta-analyses., Parkinsonism Relat. Disord. 23 (2016) 1–9. doi:10.1016/j.parkreldis.2015.12.008.

[17] L. Belbasis, V. Bellou, E. Evangelou, Environmental Risk Factors and Amyotrophic Lateral Sclerosis: An Umbrella Review and Critical Assessment of Current Evidence from Systematic Reviews and Meta-Analyses of Observational Studies., Neuroepidemiology. 46 (2016) 96–105. doi:10.1159/000443146.

[18] R. DerSimonian, N. Laird, Meta-analysis in clinical trials., Control. Clin. Trials. 7 (1986) 177–88.

[19] J. Lau, J.P. Ioannidis, C.H. Schmid, Quantitative synthesis in systematic reviews., Ann. Intern. Med. 127 (1997) 820–6.

[20] J.P.T. Higgins, S.G. Thompson, D.J. Spiegelhalter, A re-evaluation of random-effects meta-analysis., J. R. Stat. Soc. Ser. A. Stat. Soc. 172 (2009) 137–159. doi:10.1111/j.1467-985X.2008.00552.x.

[21] J.P.T. Higgins, Commentary: Heterogeneity in meta-analysis should be expected and appropriately quantified., Int. J. Epidemiol. 37 (2008) 1158–60. doi:10.1093/ije/dyn204.

[22] W.G. Cochran, The Combination of Estimates from Different Experiments, Biometrics. 10 (1954) 101. doi:10.2307/3001666.

[23] J.P.T. Higgins, S.G. Thompson, Quantifying heterogeneity in a meta-analysis., Stat. Med. 21 (2002) 1539–58. doi:10.1002/sim.1186.

[24] J.A.C. Sterne, A.J. Sutton, J.P.A. Ioannidis, N. Terrin, D.R. Jones, J. Lau, et al., Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials., BMJ. 343 (2011) d4002.

[25] M. Egger, G. Davey Smith, M. Schneider, C. Minder, Bias in meta-analysis detected by a simple, graphical test., BMJ. 315 (1997) 629–34.

[26] J.P.A. Ioannidis, T.A. Trikalinos, An exploratory test for an excess of significant findings., Clin. Trials. 4 (2007) 245–53. doi:10.1177/1740774507079441.

[27] J.P.A. Ioannidis, Clarifications on the application and interpretation of the test for excess significance and its extensions, J. Math. Psychol. 57 (2013) 184–187. doi:10.1016/j.jmp.2013.03.002.

[28] J.H. Lubin, M.H. Gail, On power and sample size for studying features of the relative

21

odds of disease., Am. J. Epidemiol. 131 (1990) 552–66.

[29] L. Cao, H. Tong, G. Xu, P. Liu, H. Meng, J. Wang, et al., Systemic Lupus Erythematous and Malignancy Risk: A Meta-Analysis, PLoS One. 10 (2015) e0122964. doi:10.1371/journal.pone.0122964.

[30] A. Green, C. Olsen, Increased Risk of Melanoma in Organ Transplant Recipients: Systematic Review and Meta-analysis of Cohort Studies, Acta Derm. Venereol. (2015). doi:10.2340/00015555-2148.

[31] A. Bauer, T.L. Diepgen, J. Schmitt, Is occupational solar ultraviolet irradiation a relevant risk factor for basal cell carcinoma? A systematic review and meta-analysis of the epidemiological literature., Br. J. Dermatol. 165 (2011) 612–25. doi:10.1111/j.1365-2133.2011.10425.x.

[32] J. Schmitt, A. Seidler, T.L. Diepgen, A. Bauer, Occupational ultraviolet light exposure increases the risk for the development of cutaneous squamous cell carcinoma: a systematic review and meta-analysis., Br. J. Dermatol. 164 (2011) 291–307. doi:10.1111/j.1365-2133.2010.10118.x.

[33] S. Gandini, F. Sera, M.S. Cattaruzza, P. Pasquini, O. Picconi, P. Boyle, et al., Meta-analysis of risk factors for cutaneous melanoma: II. Sun exposure., Eur. J. Cancer. 41 (2005) 45–60. doi:10.1016/j.ejca.2004.10.016.

[34] R.J.T. van der Leest, S.C. Flohil, L.R. Arends, E. de Vries, T. Nijsten, Risk of subsequent cutaneous malignancy in patients with prior melanoma: a systematic review and meta-analysis, J. Eur. Acad. Dermatology Venereol. 29 (2015) 1053–1062. doi:10.1111/jdv.12887.

[35] M.B. Lens, J.A. Newton-Bishop, An association between cutaneous melanoma and non-Hodgkin’s lymphoma: pooled analysis of published data with a review., Ann. Oncol. 16 (2005) 460–5. doi:10.1093/annonc/mdi080.

[36] C.M. Olsen, S.W. Lane, A.C. Green, Increased risk of melanoma in patients with chronic lymphocytic leukaemia, Melanoma Res. (2015) 1. doi:10.1097/CMR.0000000000000219.

[37] P. Huang, X.-D. Yang, S.-D. Chen, Q. Xiao, The association between Parkinson’s disease and melanoma: a systematic review and meta-analysis., Transl. Neurodegener. 4 (2015) 21. doi:10.1186/s40035-015-0044-y.

[38] C.M. Olsen, L.L. Knight, A.C. Green, Risk of melanoma in people with HIV/AIDS in the pre- and post-HAART eras: a systematic review and meta-analysis of cohort studies., PLoS One. 9 (2014) e95096. doi:10.1371/journal.pone.0095096.

[39] H. Zhao, G. Shu, S. Wang, The risk of non-melanoma skin cancer in HIV-infected patients: new data and meta-analysis, Int. J. STD AIDS. (2015). doi:10.1177/0956462415586316.

[40] C. Pouplard, E. Brenaut, C. Horreau, T. Barnetche, L. Misery, M. Richard, et al., Risk of cancer in psoriasis: a systematic review and meta-analysis of epidemiological studies, J. Eur. Acad. Dermatology Venereol. 27 (2013) 36–46. doi:10.1111/jdv.12165.

[41] M. Sanlorenzo, M.R. Wehner, E. Linos, J. Kornak, W. Kainz, C. Posch, et al., The Risk of

22

Melanoma in Airline Pilots and Cabin Crew, JAMA Dermatology. 151 (2015) 51. doi:10.1001/jamadermatol.2014.1077.

[42] A. Saxena, M. Rubens, V. Ramamoorthy, H. Khan, Risk of Second Cancers in Merkel Cell Carcinoma: A Meta-Analysis of Population Based Cohort Studies, J. Skin Cancer. 2014 (2014) 1–7. doi:10.1155/2014/184245.

[43] T.A. Simon, A. Thompson, K.K. Gandhi, M.C. Hochberg, S. Suissa, Incidence of malignancy in adult patients with rheumatoid arthritis: a meta-analysis, Arthritis Res. Ther. 17 (2015) 212. doi:10.1186/s13075-015-0728-9.

[44] S. Singh, S.J.S. Nagpal, M.H. Murad, S. Yadav, S. V Kane, D.S. Pardi, et al., Inflammatory Bowel Disease Is Associated With an Increased Risk of Melanoma: A Systematic Review and Meta-analysis, Clin. Gastroenterol. Hepatol. 12 (2014) 210–218. doi:10.1016/j.cgh.2013.04.033.

[45] X. Mariette, M. Matucci-Cerinic, K. Pavelka, P. Taylor, R. van Vollenhoven, R. Heatley, et al., Malignancies associated with tumour necrosis factor inhibitors in registries and prospective observational studies: a systematic review and meta-analysis, Ann. Rheum. Dis. 70 (2011) 1895–1904. doi:10.1136/ard.2010.149419.

[46] V.E. Johnson, Revised standards for statistical evidence., Proc. Natl. Acad. Sci. U. S. A. 110 (2013) 19313–7. doi:10.1073/pnas.1313476110.

[47] S.R. Freeman, A.L. Drake, L.F. Heilig, M. Graber, K. McNealy, L.M. Schilling, et al., Statins, Fibrates, and Melanoma Risk: a Systematic Review and Meta-analysis, JNCI J. Natl. Cancer Inst. 98 (2006) 1538–1546. doi:10.1093/jnci/djj412.

[48] A.G. Renehan, M. Tyson, M. Egger, R.F. Heller, M. Zwahlen, Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies, Lancet. 371 (2008) 569–578. doi:10.1016/S0140-6736(08)60269-X.

[49] L. Qi, X. Qi, H. Xiong, Q. Liu, J. Li, Y. Zhang, et al., Type 2 diabetes mellitus and risk of malignant melanoma: a systematic review and meta-analysis of cohort studies., Iran. J. Public Health. 43 (2014) 857–66. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4401051&tool=pmcentrez&rendertype=abstract.

[50] T.O. Yang, G.G.K. Reeves, J. Green, V. Beral, B.B.J. Cairns, H. Abbiss, et al., Birth weight and adult cancer incidence: large prospective study and meta-analysis., Ann. Oncol. 25 (2014) 1836–43. doi:10.1093/annonc/mdu214.

[51] L. Belbasis, O.A. Panagiotou, V. Dosis, E. Evangelou, A systematic appraisal of field synopses in genetic epidemiology: a HuGE review., Am. J. Epidemiol. 181 (2015) 1–16. doi:10.1093/aje/kwu249.

[52] S. Gandini, F. Sera, M.S. Cattaruzza, P. Pasquini, R. Zanetti, C. Masini, et al., Meta-analysis of risk factors for cutaneous melanoma: III. Family history, actinic damage and phenotypic factors., Eur. J. Cancer. 41 (2005) 2040–59.

[53] C.M. Olsen, H.J. Carroll, D.C. Whiteman, Estimating the Attributable Fraction for Cancer: A Meta-analysis of Nevi and Melanoma, Cancer Prev. Res. 3 (2010) 233–245. doi:10.1158/1940-6207.CAPR-09-0108.

23

[54] S. Gandini, F. Sera, M.S. Cattaruzza, P. Pasquini, D. Abeni, P. Boyle, et al., Meta-analysis of risk factors for cutaneous melanoma: I. Common and atypical naevi, Eur. J. Cancer. 41 (2005) 28–44. doi:10.1016/j.ejca.2004.10.015.

[55] M. Khalesi, D.C. Whiteman, S.A.R. Doi, J. Clark, M.G. Kimlin, R.E. Neale, Cutaneous markers of photo-damage and risk of Basal cell carcinoma of the skin: a meta-analysis., Cancer Epidemiol. Biomarkers Prev. 22 (2013) 1483–9. doi:10.1158/1055-9965.EPI-13-0424.

[56] M. Khalesi, D.C. Whiteman, B. Tran, M.G. Kimlin, C.M. Olsen, R.E. Neale, A meta-analysis of pigmentary characteristics, sun sensitivity, freckling and melanocytic nevi and risk of basal cell carcinoma of the skin., Cancer Epidemiol. 37 (2013) 534–43. doi:10.1016/j.canep.2013.05.008.

[57] F. El Ghissassi, R. Baan, K. Straif, Y. Grosse, B. Secretan, V. Bouvard, et al., A review of human carcinogens—Part D: radiation, Lancet Oncol. 10 (2009) 751–752. doi:10.1016/S1470-2045(09)70213-X.

[58] The International Agency for Research on Cancer Working Group on artificial ultraviolet (UV) light and skin cancer, The association of use of sunbeds with cutaneous malignant melanoma and other skin cancers: A systematic review, Int. J. Cancer. 120 (2006) 1116–1122. doi:10.1002/ijc.22453.

[59] S. Colantonio, M.B. Bracken, J. Beecker, The association of indoor tanning and melanoma in adults: Systematic review and meta-analysis, J. Am. Acad. Dermatol. 70 (2014) 847–857.e18. doi:10.1016/j.jaad.2013.11.050.

[60] S. Caini, M. Boniol, G. Tosti, S. Magi, M. Medri, I. Stanganelli, et al., Vitamin D and melanoma and non-melanoma skin cancer risk and prognosis: A comprehensive review and meta-analysis, Eur. J. Cancer. 50 (2014) 2649–2658. doi:10.1016/j.ejca.2014.06.024.

[61] J. Leonardi-Bee, T. Ellison, F. Bath-Hextall, Smoking and the Risk of Nonmelanoma Skin Cancer, Arch. Dermatol. 148 (2012) 939–46. doi:10.1001/archdermatol.2012.1374.

[62] C. Muranushi, C.M. Olsen, A.C. Green, N. Pandeya, Can oral nonsteroidal antiinflammatory drugs play a role in the prevention of basal cell carcinoma? A systematic review and metaanalysis., J. Am. Acad. Dermatol. 74 (2016) 108–119.e1. doi:10.1016/j.jaad.2015.08.034.

[63] M.R. Wehner, M.L. Shive, M.-M. Chren, J. Han, a. a. Qureshi, E. Linos, Indoor tanning and non-melanoma skin cancer: systematic review and meta-analysis, BMJ. 345 (2012) e5909. doi:10.1136/bmj.e5909.

[64] S. Gandini, S. Iodice, E. Koomen, A. Di Pietro, F. Sera, S. Caini, Hormonal and reproductive factors in relation to melanoma in women: Current review and meta-analysis, Eur. J. Cancer. 47 (2011) 2607–2617. doi:10.1016/j.ejca.2011.04.023.

[65] Z. Li, M. Gu, Y. Cen, Age at first birth and melanoma risk: a meta-analysis., Int. J. Clin. Exp. Med. 7 (2014) 5201–9. http://www.ncbi.nlm.nih.gov/pubmed/25664022.

[66] S. Li, Y. Liu, Z. Zeng, Q. Peng, R. Li, L. Xie, et al., Association between non-steroidal anti-inflammatory drug use and melanoma risk: a meta-analysis of 13 studies., Cancer

24

Causes Control. 24 (2013) 1505–16. doi:10.1007/s10552-013-0227-8.

[67] X. Li, X.B. Wu, Q. Chen, Statin use is not associated with reduced risk of skin cancer: a meta-analysis, Br. J. Cancer. 110 (2014) 802–807. doi:10.1038/bjc.2013.762.

[68] Z. Li, Z. Wang, Y. Yu, H. Zhang, L. Chen, Smoking is Inversely Related to Cutaneous Malignant Melanoma -Says Results from a Meta-analysis, Br. J. Dermatol. (2015). doi:10.1111/bjd.13998.

[69] M. Rota, E. Pasquali, R. Bellocco, V. Bagnardi, L. Scotti, F. Islami, et al., Alcohol drinking and cutaneous melanoma risk: a systematic review and dose-risk meta-analysis, Br. J. Dermatol. 170 (2014) 1021–1028. doi:10.1111/bjd.12856.

[70] T.N. Sergentanis, A.G. Antoniadis, H.J. Gogas, C.N. Antonopoulos, H.-O. Adami, A. Ekbom, et al., Obesity and risk of malignant melanoma: A meta-analysis of cohort and case–control studies, Eur. J. Cancer. 49 (2013) 642–657. doi:10.1016/j.ejca.2012.08.028.

[71] J. Wang, X. Li, D. Zhang, Coffee consumption and the risk of cutaneous melanoma: a meta-analysis, Eur. J. Nutr. (2015). doi:10.1007/s00394-015-1139-z.

[72] F. Xie, T. Xie, Q. Song, S. Xia, H. Li, Analysis of association between sunscreens use and risk of malignant melanoma., Int. J. Clin. Exp. Med. 8 (2015) 2378–84. http://www.ncbi.nlm.nih.gov/pubmed/25932176.

[73] Y.-P. Zhang, R.-X. Chu, H. Liu, Vitamin A Intake and Risk of Melanoma: A Meta-Analysis, PLoS One. 9 (2014) e102527. doi:10.1371/journal.pone.0102527.

[74] J. Ariyaratnam, V. Subramanian, Association between thiopurine use and nonmelanoma skin cancers in patients with inflammatory bowel disease: a meta-analysis., Am. J. Gastroenterol. 109 (2014) 163–9. doi:10.1038/ajg.2013.451.

[75] R. Liu, X. Gao, Y. Lu, H. Chen, Meta-analysis of the relationship between Parkinson disease and melanoma, Neurology. 76 (2011) 2002–2009. doi:10.1212/WNL.0b013e31821e554e.

[76] X. Mariette, A.V. Reynolds, P. Emery, Updated meta-analysis of non-melanoma skin cancer rates reported from prospective observational studies in patients treated with tumour necrosis factor inhibitors., Ann. Rheum. Dis. 71 (2012) e2. doi:10.1136/annrheumdis-2012-202478.

[77] J. Chahoud, A. Semaan, Y. Chen, M. Cao, A.G. Rieber, P. Rady, et al., Association Between β-Genus Human Papillomavirus and Cutaneous Squamous Cell Carcinoma in Immunocompetent Individuals-A Meta-analysis., JAMA Dermatology. (2015). doi:10.1001/jamadermatol.2015.4530.

[78] C. Muranushi, C.M. Olsen, N. Pandeya, A.C. Green, Aspirin and Nonsteroidal Anti-Inflammatory Drugs Can Prevent Cutaneous Squamous Cell Carcinoma: a Systematic Review and Meta-Analysis, J. Invest. Dermatol. 135 (2015) 975–983. doi:10.1038/jid.2014.531.

25

Figure 1. Flow chart of literature search

26

Table 1. Quantitative synthesis and assessment of bias across the 85 associations of non-genetic risk factors for melanoma and

keratinocyte skin cancers

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Basal cell carcinoma

Bauer, 2015[31]

Occupational

ultraviolet light

exposure

NA/NA 23 RRHigh level vs. Low

level1.44 (1.22-1.70) 1.6 × 10-5 0.74-2.81 82.7 Yes/NA

Caini, 2014[60]Serum vitamin

D1193/8227 5 OR

High level vs. Low

level1.82 (1.49-2.21) 2.7 × 10-9 1.32-2.50 0 No/No

Khalesi, 2013[55]Actinic

keratosis1709/4427 7

OR Diseased vs.

Healthy3.19 (2.57-3.95) 6.2 × 10-26 1.93-5.25 33.5 No/No

Khalesi, 2013[55] Solar elastosis 982/3715 5OR Diseased vs.

Healthy1.13 (0.52-2.45) 0.764 0.06-20.31 89.7 No/No

Khalesi, 2013[55] Solar lentigines 1709/4427 7OR Diseased vs.

Healthy1.71 (1.17-2.48) 0.005 0.49-5.99 85.2 Yes/Yes

Khalesi, 2013[55] Telangiectasia 603/3146 3OR Diseased vs.

Healthy1.59 (1.32-1.93) 2.1 × 10-6 0.46-5.55 0 No/No

Khalesi, 2013[56] Hair color 17008/435304 13 ORRed color vs. Dark

color2.02 (1.68-2.44) 2.3 × 10-13 1.31-3.12 32.2 No/No

Khalesi, 2013[56] Eye color 9539/334413 18 ORBlue color vs.

Dark color1.68 (1.37-2.07) 9.0 × 10-7 0.74-3.81 84.2 Yes/Yes

27

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Khalesi, 2013[56] Skin color 3927/7327 12 ORFair colour vs.

Dark colour2.11 (1.56-2.86) 1.3 × 10-6 0.74-6.05 83 Yes/Yes

Khalesi, 2013[56] Sunburns 4435/7039 11 ORBurn never tan vs.

Tan never burn2.03 (1.73-2.38) 2.3 × 10-18 1.27-3.26 94.4 No/No

Khalesi, 2013[56]Freckles in

childhood1790/4369 8 OR

Exposed vs. Not

exposed1.57 (1.29-1.92) 9.9 × 10-6 0.93-2.65 49.5 Yes/Yes

Leonardi,

2012[61]Smoking 8971/184880 17 OR

Exposed vs. Not

exposed0.95 (0.82-1.09) 0.445 0.60-1.49 58.9 No/No

Muranushi,

2016[62]Aspirin 101257/378807 8 RR

Exposed vs. Not

exposed0.95 (0.91-0.99) 0.030 0.85-1.07 55.1 Yes/No

Muranushi,

2016[62]

Non-aspirin

NSAIDs91011/317805 7 RR

Exposed vs. Not

exposed0.94 (0.87-1.02) 0.156 0.74-1.20 84.3 No/No

Muranushi,

2016[62]

Aspirin or non-

aspirin NSAIDs107429/395816 11 RR

Exposed vs. Not

exposed0.91 (0.85-0.98) 8.0 × 10-3 0.74-1.12 85.3 No/Yes

Wehner, 2012[63] Indoor tanning 7375/70603 8 ORExposed vs. Not

exposed1.29 (1.08-1.53) 4.4 × 10-3 0.86-1.92 36.7 No/No

Cutaneous melanoma

Caini, 2014[60]Serum vitamin

D392/11573 4 OR

High levels vs.

Low levels1.56 (0.80-3.04) 0.192 0.12-20.28 54.4 Yes/No

Caini, 2014[60]Vitamin D

intake1678/105755 5 RR

High intake vs.

Low intake0.87 (0.66-1.16) 0.349 0.35-2.20 64.6 No/No

Cao, 2015[29]Systemic lupus

erythematosus56/NA 5 SIR

Diseased vs.

Healthy0.65 (0.50-0.85) 1.7 × 10-3 0.43-1.01 0 No/NA

28

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Colantonio,

2014[59]Indoor tanning 14956/233106 31 OR Ever vs. Never use 1.16 (1.06-1.28) 2.4 × 10-3 0.80-1.68 51.3 No/No

Freeman,

2006[47]Fibrates 27/15530 6 OR Ever vs. Never use 0.56 (0.22-1.42) 0.222 0.13-2.49 4.4 No/No

Gandini, 2005[33]Chronic sun

exposure33634/NA 41 OR

High level vs. Low

level0.95 (0.88-1.04) 0.281 0.61-1.36 58.9 No/NA

Gandini, 2005[52]Density of

freckles10103/12117 33 OR

High level vs. Low

level2.10 (1.80-2.45) 5.7 × 10-21 1.03-4.29 65.7 No/No

Gandini, 2005[52] Eye colour 7930/12273 34 ORLight colour vs.

Dark colour1.61 (1.44-1.81) 7.4 × 10-16 0.96-2.71 57.3 Yes/No

Gandini, 2005[52] Hair colour 9473/13907 39 ORLight colour vs.

Dark colour1.77 (1.62-1.94) 1.1 × 10-35 1.22-2.58 43.2 No/No

Gandini, 2005[52]Indicators of

actinic damage3804/4328 10 OR

Diseased vs.

Healthy2.00 (1.23-3.27) 5.5 × 10-3 0.36-11.01 87.8 No/No

Gandini, 2005[33]Intermittent sun

exposure8461/12576 33 OR

High level vs. Low

level1.62 (1.31-1.99) 7.2 × 10-6 0.54-4.87 82.7 Yes/Yes

Gandini, 2005[52]Premalignant

skin lesion2810/3809 11 OR

Diseased vs.

Healthy4.24 (2.77-6.50) 2.9 × 10-11 1.01-17.78 77.8 No/No

Gandini, 2005[52] Skin colour 20823/35536 31 OR

Light colour vs.

Medium or dark

colour

2.05 (1.67-2.51) 3.6 × 10-12 0.76-5.51 83.4 Yes/Yes

Gandini, 2005[52] Skin type 6942/9139 22 OR Type I vs. Type IV 2.08 (1.69-2.57) 1.1 × 10-11 0.99-4.41 54.4 Yes/No

Gandini, 2005[33] Sunburns 8256/11306 33 ORHigh level vs. Low

level2.02 (1.73-2.37) 1.1 × 10-18 0.98-4.17 62.4 Yes/Yes

29

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Gandini, 2005[33]Total sun

exposure3906/6253 15 OR

High level vs. Low

level1.33 (1.01-1.76) 0.041 0.47-3.80 79.6 No/No

Gandini, 2011[64]

Hormone

replacement

therapy

2816/107266 10 RRExposed vs. Not

exposed1.16 (0.96-1.41) 0.135 0.68-1.98 48.3 No/No

Gandini, 2011[64]Oral

contraceptives4764/296583 20 RR

Exposed vs. Not

exposed1.00 (0.84-1.18) 0.955 0.52-1.89 67.4 No/No

Gandini, 2011[64] Parity 16197/3039786 18 RRPer 1 child

increase0.97 (0.93-1.01) 0.174 0.84-1.12 67.7 No/No

Green, 2015[30]Organ

transplantation927/NA 20 SIR

Exposed vs. Not

exposed2.71 (2.23-3.30) 1.6 × 10-23 1.38-5.34 78.1 No/NA

Huang, 2015[37]Parkinson’s

diseaseNA/NA 14 RR

Diseased vs.

Healthy2.43 (1.77-3.32) 3.2 × 10-8 0.77-7.66 87.8 Yes/NA

Leest, 2015[34]History of

melanoma6727/NA 12 SIR

Diseased vs.

Healthy12.78 (9.34-17.49) 4.8 × 10-57 3.88-42.04 98.6 No/NA

Lens, 2005[35]Non Hodgkin

lymphoma310/NA 7 SIR

Diseased vs.

Healthy2.14 (1.81-2.53) 6.5 × 10-19 1.47-3.12 31.5 No/NA

Li, 2014[65]Age at first

birth14443/3144199 11 RR

Oldest vs.

Youngest age1.45 (1.05-2.02) 0.025 0.47-4.49 85.6 No/No

Li, 2013[66] Aspirin 7804/897634 9 RRExposed vs. Not

exposed0.97 (0.85-1.09) 0.582 0.66-1.41 70.5 No/No

Li, 2013[66]Aspirin or non-

aspirin NSAIDs11011/940980 12 RR

Exposed vs. Not

exposed0.97 (0.89-1.05) 0.428 0.78-1.20 46.8 No/No

30

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Li, 2013[66]Non-aspirin

NSAIDs6919/752406 8 RR

Exposed vs. Not

exposed0.98 (0.88-1.08) 0.635 0.74-1.29 59.1 No/No

Li, 2014[67] Statins 10113/613124 8 RRExposed vs. Not

exposed0.93 (0.84-1.03) 0.152 0.71-1.22 34.9 No/No

Li, 2015[68] Smoking 7300/685267 20 RRExposed vs. Not

exposed0.86 (0.80-0.92) 1.5 × 10-5 0.72-1.03 33.4 Yes/Yes

Olsen, 2010[53] Atypical nevi 6086/7427 26 ORAt least 1 nevus

vs. None nevi3.63 (2.85-4.62) 1.1 × 10-25 1.19-11.07 77.7 No/No

Olsen, 2010[53] Common nevi 5560/6994 24 OR Per 1 nevi increase 1.017 (1.01-1.02) 1.9 × 10-29 1.01-1.03 76.5 No/Yes

Olsen, 2014[38] AIDS 562/NA 26 SIRDiseased vs.

Healthy1.25 (1.08-1.45) 2.8 × 10-3 0.81-1.94 35.9 No/NA

Olsen, 2015[36]

Chronic

lymphocytic

leukemia

443/NA 7 SIRDiseased vs.

Healthy3.88 (2.08-7.22) 1.9 × 10-5 0.46-32.65 96 No/NA

Pouplard,

2013[40]Psoriasis 74/NA 6 SIR

Diseased vs.

Healthy1.16 (0.81-1.65) 0.414 0.48-2.80 38.7 No/NA

Qi, 2014[49]Type 2 diabetes

mellitus5316/5717963 9 RR

Diseased vs.

Healthy1.15 (1.00-1.32) 0.046 0.80-1.66 57.6 No/No

Renehan,

2008[48]BMI 8278/3968582 11 RR

Per 5 kg/m2

increase1.07 (0.97- 1.18) 0.174 0.80-1.43 77.3 No/No

Rota, 2014[69]Alcohol

drinking6251/1351482 17 RR

Exposed vs. Not

exposed1.20 (1.06-1.37) 5.8 × 10-3 0.81-1.80 55.7 No/No

Sanlorenzo,

2015[41]

Airline pilots

and cabin crewNA/NA 14 SIR

Exposed vs. Not

exposed2.21 (1.76-2.77) 7.4 × 10-12 1.09-4.47 64.7 No/NA

31

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Saxena, 2014[42]Merkell cell

carcinoma46/NA 3 SIR

Diseased vs.

Healthy3.09 (2.02-4.73) 2.0 × 10-7 0.20-48.67 0 No/NA

Sergentanis,

2013[70]BMI in men 5661/4983231 15 OR

Obese vs. Normal

weight1.33 (1.14-1.55) 3.6 × 10-4 0.95-1.85 23 No/No

Sergentanis,

2013[70]BMI in women 3525/1542335 16 OR

Obese vs. Normal

weight0.99 (0.83-1.19) 0.942 0.61-1.61 43.3 No/No

Sergentanis,

2013[70]BMI in men 1601/1259709 13 OR

Overweight vs.

Normal weight1.36 (1.16-1.59) 1.8 × 10-4 0.98-1.89 21.8 No/No

Sergentanis,

2013[70]BMI in women 3795/1576562 12 OR

Overweight vs.

Normal weight0.96 (0.88-1.05) 0.410 0.87-1.07 0 No/No

Sergentanis,

2013[70]BSA in men 566/695 5 OR ≥2 m2 vs. <2 m2 1.74 (1.02-2.98) 0.044 0.35-8.75 53.1 No/No

Sergentanis,

2013[70]BSA in women 3755/111287 7 OR ≥2 m2 vs. <2 m2 1.37 (0.94-2.00) 0.099 0.53-3.55 46.1 No/No

Simon, 2015[43]Rheumatoid

arthritis674/NA 21 SIR

Diseased vs.

Healthy1.21 (1.00-1.46) 0.048 0.60-2.44 80.7 No/NA

Singh, 2014[44]Inflammatory

bowel disease179/NA 12 SIR

Diseased vs.

Healthy1.37 (1.10- 1.70) 4.3 × 10-3 0.84-2.22 32.5 No/NA

Wang, 2015[71] Coffee drinking 7140/825589 12 RRHigh intake vs.

Low intake0.80 (0.70-0.93) 3.0 × 10-3 0.53-1.21 53.5 No/No

Xie, 2015[72] Sunscreen use 7522/15970 21 RRExposed vs. Not

exposed1.14 (0.91-1.44) 0.249 0.43-3.03 83.6 No/Yes

Yang, 2015[50] Birth weight 4000/3821122 6 RR Per 1 kg increase 1.14 (1.05-1.24) 1.9 × 10-3 1.01-1.29 0 No/No

32

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Zhang, 2014[73] Retinol intake 2776/231607 9 RRHigh intake vs.

Low intake0.80 (0.69-0.93) 2.8 × 10-3 0.67-0.95 0 No/No

Zhang, 2014[73]Vitamin A

intake1912/69859 5 RR

High intake vs.

Low intake0.86 (0.59-1.25) 0.421 0.26-2.86 66.8 No/No

Zhang, 2014[73]β-carotene

intake2613/231656 9 RR

High intake vs.

Low intake0.87 (0.62-1.20) 0.392 0.31-2.40 71.9 No/No

Keratinocyte skin cancers

Ariyaratnam,

2014[74]

Thiopurine use

in patients with

inflammatory

bowel disease

14081/46000 8 HRExposed vs. Not

exposed2.28 (1.50-3.45) 1.1 × 10-4 0.67-7.71 76.9 No/No

Caini, 2014[60]Vitamin D

intake7408/145813 4 RR

High intake vs.

Low intake1.03 (0.98-1.09) 0.216 0.92-1.16 0 No/No

Cao, 2015[29]Systemic lupus

erythematosus8/NA 3 SIR

Diseased vs.

Healthy1.50 (1.06-2.12) 0.021 0.08- 27.89 23.1 No/NA

Li, 2014[67] Statins 3578/164218 19 RRExposed vs. Not

exposed0.93 (0.84-1.03) 0.152 0.71-1.22 34.9 No./No

Liu, 2011[75]Parkinson’s

disease2114/11254312 9 OR

Diseased vs.

Healthy1.14 (0.99-1.32) 0.065 0.81-1.61 49.7 No/No

Mariette,

2012[76]TNF inhibitors 833/NA 5 SIR

Exposed vs. Not

exposed1.39 (1.14-1.68) 8.6 × 10-4 1.01-1.89 0 No/NA

Zhao, 2014[39] AIDS 4314/NA 6 SIRDiseased vs.

Healthy3.46 (1.59-7.54) 1.8 × 10-3 0.20-60.20 98.3 No/NA

Squamous cell carcinoma

33

Reference Risk factorNumber of

cases/controls

Number of

datasets*

Effect

size

Level of

comparison

Random-effects

summary effect

size (95% CI)

P random95% prediction

intervalI2

Small-study

effects/Excess

statistical

significance

Caini, 2014[60]Serum vitamin

D300/8245 4 RR

High levels vs.

Low levels1.80 (0.64-5.04) 0.266 0.02-155.62 81.4 No/No

Chahoud,

2016[77]

β-genus HPV

infection2199/3836 9 OR

Diseased vs.

Healthy1.42 (1.18-1.72) 2.8 × 10-4 0.88-2.30 45.1 No/Yes

Leonardi,

2012[61]Smoking 1591/109883 7 OR

Exposed vs. Not

exposed1.52 (1.15-2.01) 3.3 × 10-3 0.66-3.50 64.4 No/Yes

Muranushi,

2015[78]Aspirin 4663/112826 6 RR

Exposed vs. Not

exposed0.88 (0.75-1.03) 0.098 0.55-1.39 63.7 No/No

Muranushi,

2015[78]

Aspirin or non-

aspirin NSAIDs4881/114032 8 RR

Exposed vs. Not

exposed0.84 (0.73-0.96) 0.010 0.56-1.24 65 No/No

Muranushi,

2015[78]

Non-aspirin

NSAIDs4449/97096 6 RR

Exposed vs. Not

exposed0.86 (0.78-0.95) 2.0 × 10-3 0.75-0.98 0 No/No

Pouplard,

2013[40]Psoriasis 351/NA 7 SIR

Diseased vs.

Healthy5.31 (2.63-10.71) 3.2 × 10-6 0.46-61.23 95.4 No/NA

Schmitt, 2011[32]

Occupational

ultraviolet light

exposure

NA/NA 18 RRHigh level vs. Low

level1.76 (1.42-2.18) 3.3 × 10-7 0.84-3.69 75.7 Yes/NA

Wehner, 2012[63] Indoor tanning 1635/75335 6 ORExposed vs. Not

exposed1.66 (1.29-2.14) 7.1 × 10-5 0.86-3.21 45.5 No/No

* The number of datasets does not always overlap with the number of studies. In some meta-analyses, a component study is separated

into two datasets, using sex-specific effect sizes for males and females.

34

AIDS: acquired immune deficiency syndrome, BMI: body mass index, BSA: body surface area, CI: confidence interval, HPV: human

papilloma virus, HR: hazard ratio, NA: not available, NP: not pertinent because the number of expected significant studies was larger

than the number of observed significant studies, NSAIDs: non-steroid anti-inflammatory drugs, OR: odds ratio, RR: risk ratio, SIR:

standardized incidence ratio, TNF: tumor necrosis factor

35

Table 2. Assessment across the statistically significant associations for melanoma and keratinocyte skin cancers

Level of

evidence

Criteria Cutaneous melanoma Keratinocyte

skin cancers

Basal cell

carcinoma

Squamous cell

carcinoma

Convincing >1000 cases, P<10-6, I2<50%,

95% prediction interval

excluding the null value, no

evidence for small-study

effects and excess

significance bias

Hair color None Actinic

keratosis, serum

vitamin D, hair

color

None

Highly

suggestive

>1000 cases, P<10-6, largest

study with a statistically

significant effect

Atypical nevi, density of freckles, eye color, skin

type, sunburns, premalignant skin lesions, common

nevi, history of melanoma

None Sunburns None

Suggestive >1000 cases, P<10-3 BMI in men (overweight vs. normal weight and

obese vs. normal weight), intermittent sun

exposure, skin color, smoking

Thiopurine use

in IBD patients

Eye color,

freckles in

childhood,

occupational

ultraviolet light

exposure, skin

color

Indoor tanning,

occupational

ultraviolet light

exposure, β-genus

HPV infection

36

Level of

evidence

Criteria Cutaneous melanoma Keratinocyte

skin cancers

Basal cell

carcinoma

Squamous cell

carcinoma

Weak AIDS, airline pilots and cabin crew, age at first

birth, alcohol drinking, birth weight, BSA in men

(≥2 vs. <2), coffee drinking, CLL, indicators of

actinic damage, indoor tanning, non-Hodgkin

lymphoma, systemic lupus erythematosus, total sun

exposure, organ transplantation, Type 2 diabetes

mellitus, history of Merkel cell carcinoma,

rheumatoid arthritis, Parkinson’s disease, IBD,

retinol intake

AIDS,

systemic lupus

erythematosus,

TNF inhibitors

Aspirin, aspirin

or non-aspirin

NSAIDs, indoor

tanning, solar

lentigines,

telangiectasia

Smoking, aspirin

or any non-aspirin

NSAIDs, non-

aspirin NSAIDs,

psoriasis

37

Supplementary Table 1. Heterogeneity estimates, bias assessment and largest study effect size across the 85 associations of non-

genetic risk factors for melanoma and keratinocyte skin cancers

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Basal cell carcinoma

Bauer,

2015[31]

Occupational

ultraviolet

light exposure

RR 1.00 (0.90-1.10) 0.051 82.7 0.021 10 NA NA

Caini,

2014[60]

Serum vitamin

DOR 1.51 (1.10-2.07) 0.161 0 0.610 3 4.06 NP

Khalesi,

2013[55]

Actinic

keratosis

OR2.98 (2.27-3.91) 0.139 33.5 0.696 7 6.99 0.983

Khalesi,

2013[55]Solar elastosis

OR1.70 (1.29-2.25) 0.142 89.7 0.344 3 4.43 NP

Khalesi,

2013[55]

Solar

lentigines

OR1.28 (1.01-1.63) 0.122 85.2 0.084 6 2.68 0.010

Khalesi,

2013[55]Telangiectasia

OR1.57 (1.21-2.04) 0.133 0 0.612 2 2.52 NP

Khalesi,

2013[56]Hair colour

OR2.00 (1.69-2.38) 0.087 32.2 0.137 8 11.65 NP

Khalesi,

2013[56]Eye colour

OR1.10 (0.98-1.22) 0.056 84.2 0.034 10 2.98 8.4 × 10-6

Khalesi,

2013[56]Skin colour

OR1.18 (1.04-1.34) 0.065 83 0.002 8 2.91 6.1 × 10-4

38

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Khalesi,

2013[56]Sunburns

OR2.27 (2.20-2.33) 0.015 94.4 0.717 9 10.65 NP

Khalesi,

2013[56]

Freckles in

childhood

OR1.23 (1.08-1.40) 0.066 49.5 0.007 6 2.23 0.003

Leonardi,

2012[61]Smoking OR 0.90 (0.81-1.00) 0.054 58.9 0.455 5 3.25 0.280

Muranushi,

2016[62]Aspirin RR 0.99 (0.95-1.02) 0.018 55.1 0.012 1 0.63 0.630

Muranushi,

2016[62]

Non-aspirin

NSAIDsRR 0.96 (0.92-1.00) 0.021 84.3 0.531 2 2.57 0.654

Muranushi,

2016[62]

Aspirin or

non-aspirin

NSAIDs

RR 1.00 (0.98-1.03) 0.013 85.3 0.147 2 0.55 0.045

Wehner,

2012[63]Indoor tanning OR 1.29 (1.22-1.35) 0.026 36.7 0.510 3 3.73 NP

Cutaneous melanoma

Caini,

2014[60]

Serum vitamin

DOR 0.82 (0.44-1.55) 0.321 54.4 0.063 0 0.55 NP

Caini,

2014[60]

Vitamin D

intakeRR 1.13 (0.89-1.43) 0.121 64.6 0.338 1 0.92 0.930

Cao, 2015[29]Systemic lupus

erythematosusSIR 0.67 (0.48-0.93) 0.169 0 0.955 1 NA NA

Colantonio,

2014[59]Indoor tanning OR 1.11 (0.97-1.27) 0.069 51.3 0.678 7 5.00 0.328

39

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Freeman,

2006[47]Fibrates OR 0.33 (0.07-1.64) 0.804 4.4 0.748 0 1.50 NP

Gandini,

2005[33]

Chronic sun

exposureOR 0.86 (0.81-0.92) 0.032 58.9 0.214 11 NA NA

Gandini,

2005[52]

Density of

frecklesOR 1.59 (1.29-2.00) 0.112 65.7 0.102 26 24.13 0.464

Gandini,

2005[52]Eye colour OR 1.60 (1.24-2.06) 0.129 57.3 0.090 18 24.06 NP

Gandini,

2005[52]Hair colour OR 1.54 (1.23-1.93) 0.115 43.2 0.494 27 26.03 0.741

Gandini,

2005[52]

Indicators of

actinic damageOR 1.27 (0.99-1.68) 0.135 87.8 0.223 7 4.46 0.106

Gandini,

2005[33]

Intermittent

sun exposureOR 1.04 (0.82-1.33) 0.123 82.7 0.064 19 1.91 1.0 × 10-8

Gandini,

2005[52]

Premalignant

skin lesionOR 2.61 (2.21-3.14) 0.090 77.8 0.155 10 8.40 0.256

Gandini,

2005[52]Skin colour OR 1.00 (0.91-1.11) 0.051 83.4 1.2 × 10-5 19 1.55 1.0 × 10-8

Gandini,

2005[52]Skin type OR 1.49 (1.11-2.00) 0.150 54.4 0.032 13 15.65 NP

Gandini,

2005[33]Sunburns OR 1.39 (1.05-1.88) 0.149 62.4 0.001 24 17.02 0.015

Gandini,

2005[33]

Total sun

exposureOR 1.32 (1.05-1.69) 0.121 79.6 0.775 9 6.65 0.216

40

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Gandini,

2011[64]

Hormone

replacement

therapy

RR 0.90 (0.72-1.13) 0.115 48.3 0.544 2 1.32 0.527

Gandini,

2011[64]

Oral

contraceptivesRR 1.30 (1.07-1.55) 0.095 67.4 0.806 3 8.26 NP

Gandini,

2011[64]Parity RR 0.91 (0.89-0.94) 0.014 67.7 0.219 3 3.99 NP

Green,

2015[30]

Organ

transplantationSIR 2.38 (2.14-2.63) 0.053 78.1 0.644 10 NA NA

Huang,

2015[37]

Parkinson’s

diseaseRR 1.19 (1.04-1.36) 0.068 87.8 0.009 11 NA NA

Leest,

2015[34]

History of

melanomaSIR 8.60 (8.30-8.90) 0.018 98.6 0.266 11 NA NA

Lens,

2005[35]

Non Hodgkin

lymphomaSIR 1.75 (1.47-2.07) 0.087 31.5 0.567 5 NA NA

Li, 2014[65]Age at first

birthRR 1.37 (1.17-1.61) 0.081 85.6 0.863 7 7.09 NP

Li, 2013[66] Aspirin RR 0.89 (0.80-0.98) 0.052 70.5 0.684 4 3.28 0.618

Li, 2013[66]

Aspirin or

non-aspirin

NSAIDs

RR 0.87 (0.80-0.94) 0.041 46.8 0.831 3 5.13 NP

Li, 2013[66]Non-aspirin

NSAIDsRR 0.85 (0.79-0.91) 0.036 59.1 0.265 1 4.04 NP

Li, 2014[67] Statins RR 0.89 (0.75-1.06) 0.088 34.9 0.878 3 4.89 NP

Li, 2015[68] Smoking RR 1.02 (0.95-1.10) 0.037 33.4 0.005 4 1.09 0.004

41

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Olsen,

2010[53]Atypical nevi OR 2.48 (1.82-3.38) 0.158 77.7 0.310 21 25.75 NP

Olsen,

2010[53]Common nevi OR 1.019 (1.016-1.023) 0.002 76.5 0.112 20 1.24 1.0 × 10-8

Olsen,

2014[38]AIDS SIR 1.30 (1.10-1.53) 0.084 35.9 0.449 5 NA NA

Olsen,

2015[36]

Chronic

lymphocytic

leukemia

SIR 7.74 (6.85-8.72) 0.062 96 0.358 6 NA NA

Pouplard,

2013[40]Psoriasis SIR 0.95 (0.69-1.30) 0.162 38.7 0.219 1 NA NA

Qi, 2014[49]

Type 2

diabetes

mellitus

RR 1.16 (1.07-1.27) 0.044 57.6 0.392 4 2.36 0.213

Renehan,

2008[48]BMI RR 0.93 (0.87-0.99) 0.033 77.3 0.359 4 2.11 0.147

Rota, 2014[69]Alcohol

drinkingRR 1.26 (1.18-1.34) 0.032 55.7 0.991 5 6.01 NP

Sanlorenzo,

2015[41]

Airline pilots

and cabin crewSIR 3.47 (2.85-4.22) 0.100 64.7 0.695 10 NA NA

Saxena,

2014[42]

Merkell cell

carcinomaSIR 3.31 (1.89-5.37) 0.266 0 0.457 2 NA NA

42

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Sergentanis,

2013[70]

BMI in men

(obese vs.

normal

weight)

OR 1.27 (1.12-1.45) 0.066 21.8 0.294 4 2.84 0.437

Sergentanis,

2013[70]

BMI in

women (obese

vs. normal

weight)

OR 0.99 (0.86-1.13) 0.071 0 0.163 1 0.61 0.608

Sergentanis,

2013[70]

BMI in men

(overweight

vs. normal

weight)

OR 1.29 (1.14-1.46) 0.063 23 0.802 5 4.24 0.664

Sergentanis,

2013[70]

BMI in

women

(overweight

vs. normal

weight)

OR 0.94 (0.83-1.07) 0.065 43.3 0.300 1 1.14 NP

Sergentanis,

2013[70]BSA in men OR 2.05 (1.31-3.21) 0.227 53.1 0.781 2 3.48 NP

Sergentanis,

2013[70]

BSA in

womenOR 1.60 (1.03-2.48) 0.224 46.1 0.978 2 3.88 NP

Simon,

2015[43]

Rheumatoid

arthritisSIR 1.47 (1.31-1.65) 0.059 80.7 0.959 7 NA NA

Singh,

2014[44]

Inflammatory

bowel diseaseSIR 1.29 (1.09-1.53) 0.087 32.5 0.596 3 NA NA

43

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Wang,

2015[71]

Coffee

drinkingRR 0.80 (0.68-0.93) 0.080 53.5 0.738 5 6.46 NP

Xie, 2015[72] Sunscreen use RR 0.82 (0.70-0.96) 0.081 83.6 0.194 12 7.01 0.021

Yang,

2015[50]Birth weight RR 1.13 (1.00-1.27) 0.061 0 0.912 2 2.25 NP

Zhang,

2014[73]Retinol intake RR 0.84 (0.64-1.10) 0.138 0 0.306 2 2.50 NP

Zhang,

2014[73]

Vitamin A

intakeRR 0.87 (0.66-1.13) 0.137 66.8 0.879 1 1.12 NP

Zhang,

2014[73]

β-carotene

intakeRR 1.13 (0.86-1.49) 0.140 71.9 0.672 2 1.45 0.616

Keratinocyte skin cancers

Ariyaratnam,

2014[74]

Thiopurine use

in patients

with

inflammatory

bowel disease

HR 2.28 (1.86-2.79) 0.103 76.9 0.847 5 7.84 NP

Caini,

2014[60]

Vitamin D

intakeRR 1.02 (0.95-1.07) 0.030 0 0.114 0 0.30 NP

Cao, 2015[29]Systemic lupus

erythematosusSIR 1.53 (0.98-2.28) 0.215 23.1 0.731 1 NA NA

Li, 2014[67] Statins RR 0.89 (0.75-1.06) 0.088 34.9 0.878 3 4.89 NP

Liu, 2011[75]Parkinson’s

diseaseOR 1.25 (1.10-1.40) 0.062 49.7 0.556 3 2.54 0.731

44

Reference Risk factorEffect size

metric

Largest study effect

size (95% CI)SE I2 (%)

Egger test p-

value

Observed

significant

studies

Expected

significant

studies

Excess

significance

test p-value

Mariette,

2012[76]TNF inhibitors SIR 1.51 (1.20-1.95) 0.124 0 0.493 1 NA NA

Zhao,

2014[39]AIDS SIR 2.10 (1.91-2.31) 0.049 98.3 0.507 6 NA NA

Squamous cell carcinoma

Caini,

2014[60]

Serum vitamin

DRR 0.67 (0.44-1.03) 0.217 81.4 0.220 1 1.63 NP

Chahoud,

2016[77]

β-genus HPV

infectionOR 1.30 (1.05-1.62) 0.111 45.1 0.112 6 3.36 0.068

Leonardi,

2012[61]Smoking OR 1.18 (0.95-1.48) 0.113 64.4 0.370 4 1.37 0.012

Muranushi,

2015[78]Aspirin RR 0.86 (0.76-0.98) 0.065 63.7 0.818 2 2.66 NP

Muranushi,

2015[78]

Aspirin or

non-aspirin

NSAIDs

RR 0.85 (0.76-0.94) 0.054 65 0.679 2 3.11 NP

Muranushi,

2015[78]

Non-aspirin

NSAIDsRR 0.85 (0.75-0.97) 0.066 0 0.598 1 2.79 NP

Pouplard,

2013[40]Psoriasis SIR 11.90 (10.10-14.02) 0.084 95.4 0.359 6 NA NA

Schmitt,

2011[32]

Occupational

ultraviolet

light exposure

RR 1.00 (0.90-1.10) 0.051 75.7 3.1 × 10-6 13 NA NA

Wehner,

2012[63]Indoor tanning OR 1.50 (1.20-1.78) 0.100 45.5 0.309 3 4.73 NP

45

AIDS: acquired immune deficiency syndrome, BMI: body mass index, BSA: body surface area, CI: confidence interval, HPV: human

papilloma virus, HR: hazard ratio, NA: not available, NP: not pertinent because the number of expected significant studies was larger

than the number of observed significant studies, NSAIDs: non-steroid anti-inflammatory drugs, OR: odds ratio, RR: risk ratio, SIR:

standardized incidence ratio, TNF: tumor necrosis factor

46