Abstract
Background: There is an ongoing debate whether solarium use (indoor tanning/artificial UV) may increase the risk for primary cutaneous malignant melanoma. Aim: A systematic literature search was conducted using MEDLINE and ISI Web of Science. Included studies were critically assessed regarding their risk of bias, and methodological shortcomings. Levels of evidence and grades of recommendation were determined according to guidelines of the Oxford Centre for Evidence-Based Medicine. Summary risk estimates and 95% confidence intervals for four different outcomes (ever exposure, exposure at younger age, high/low exposure vs. non-exposure) were derived from random-effects meta-analyses to account for possible heterogeneity across studies. Results: Two cohort and twenty-nine case–control studies were eligible. Overall, quality of included studies was poor as a result of severe limitations, including possible recall and selection bias, and due to lack of interventional trials. Summary risk estimates suggested a weak association (odds ratio (OR)=1.19, 95% confidence interval (CI)=1.04-1.35, p=0.009) for ever-exposure to UV radiation from a solarium with melanoma risk. However, sensitivity analyses did not show an association for studies from Europe (OR=1.10; 95%CI=0.95-1.27, p=0.218), studies with low risk of bias (OR=1.15; 95%CI=0.94-1.41, p=0.179), and studies conducted after 1990 (OR 1.09; 95%CI=0.93-1.29, p=0.295). Moreover, moderate associations were found for first exposure to UV radiation from a solarium at younger age (<25 years) and high exposure (>10 sessions in lifetime) with melanoma risk. However, for all outcomes analyzed, overall study quality and resulting levels of evidence (3a–) and grades of recommendation (D) were low due to lack of interventional studies and severe limitations including unobserved or unrecorded confounding. Conclusion: Current scientific knowledge is mainly based on observational studies with poor quality data, which report associations but do not prove causality. At present, there is no convincing evidence that moderate/responsible solarium use increases melanoma risk.
Sunlight represents an important pre-requisite for the development of life on earth and for human evolution (1, 2). Of particular importance is the ultraviolet (UV) range (UV-C: 200-280 nm; UV-B: 280-315 nm; UV-A: 315-400 nm) of solar radiation, because exposure to solar or artificial UV exerts both positive and negative effects on human health (1-5). While some of the beneficial UV effects are due to the UV-B-mediated cutaneous synthesis of vitamin D (1-5), hazardous effects include the initiation and promotion of skin photocarcinogenesis (6). The relevance of the UV-B spectrum in promoting non-melanoma skin cancer (risk factor: high cumulative exposure, via e.g. induction of DNA mutations) and melanoma (risk factor: high intermittent exposure, e.g. sunburn, most importantly in childhood) is generally accepted. Laboratory and animal studies have more recently suggested a possible additional contribution of the UV-A spectrum to skin photocarcinogenesis (5, 6). Malignant melanoma represents the most aggressive form of skin cancer. While melanoma death rates had more than doubled in light-skinned populations between 1955 and 1985, decreases in melanoma mortality rates were observed from 1985-1990 in Australia, the United States and in many European countries (7).
In this context, the beneficial and hazardous effects of solarium use, in particular the relevance of solarium use as a potential risk factor for malignant melanoma, are still a matter of debate. Since the 1980s, solarium use has been widely practised in Western and Northern Europe, Canada and the United States, and since 2000 it has become common even in sun-rich countries such as Australia (8). Modern solarium devices mainly emit UV-A radiation, only a small fraction (<5%) of radiation is emitted in the UV-B range (8). Many studies have investigated the impact of indoor tanning on melanoma risk (8-59), however, most of them have been criticized for limitations, unbalanced view, errors or incorrectnesses (11, 17). While some reports suggest that solarium use may increase melanoma risk (e.g. 19, 29, 32), other investigations found no or even a protective effect (e.g. 20-22, 24, 35). In 2009, the International Agency for Research on Cancer (IARC) classified the complete ultraviolet spectrum (including UV-A) and solarium devices as Group One carcinogens to humans (9). To improve our present understanding about the relevance of solarium use as a potential risk factor for malignant melanoma, we performed a meta-analysis and evidence-based systematic review of the literature.
Materials and Methods
This meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) (60), and the Meta-analysis of Observational studies in Epidemiology (MOOSE) guidelines (61).
Search strategy and inclusion criteria. The following terms and their variations were considered to be synonymous with solarium: indoor tanning, sunbed, sunlamp, artificial UV exposure, and non-solar UV-exposure. Relevant publications were searched independently by four Authors (BB, IH, SR, JR) in MEDLINE (from 1946) and ISI Web of Science (from 1945) using combinations of the following keywords: “sunbed”, “sunlamp”, “solarium”, “solaria”, “artificial UV”, “artificial light”, “UV”, “indoor tanning”, “tanning bed”, “tanning parlour”, “tanning salon”, “tanning booth”, “skin cancer”, and “melanoma”. Identified articles including reviews were cross-referenced for articles missed by database search. Inclusion criteria were study type/content (interventional, cohort, case–control studies published up until January 15, 2016, which reported an association between exposure to UV radiation from a solarium and cutaneous malignant melanoma), types of outcome measures (for retrospective and prospective studies: development of melanoma) and no language restrictions. In case of duplicate samples, the most recent study was included. Exclusion criteria were defined accordingly.
Outcome measures. Main outcome measures were defined as “ever-exposure to UV radiation from a solarium (yes)”, “high/low exposure to UV radiation from a solarium” and “first exposure to UV radiation from a solarium at younger age”. All exposures were compared against non-exposure. Data were independently extracted by three Authors (BB, JR and JS), and transferred to a data collection sheet that considered relevant parts of the Cochrane Consumers and Communication Review Group's data extraction template (62).
Assessment of risk of bias. Risk of bias was assessed independently by two authors (JR, JS) using a modification of the Newcastle-Ottawa Quality Assessment Scale (MNOS;) (63). Moreover, potential biases derived from methodological shortcomings were evaluated by SW and JS. Consensus regarding the grading was sought and disagreements were discussed with a third Author (JR). Finally, the risk of bias was indicated as low risk of bias (MNOS>4) and high risk of bias (MNOS≤4).
Level of evidence and grade of recommendation. For main outcomes, level of evidence and grade of recommendation were determined according to the Oxford Centre for Evidence-based Medicine (64).
Statistical analysis. Reported risk estimates [odds ratio (OR) or hazard ratio (HR)] across included studies were not consistently adjusted for confounders causing methodological heterogeneity. Accordingly, crude risk estimates (only OR) and their 95% confidence intervals (CI) were calculated from studies' contingency tables, or were taken from the original articles. Additionally, adjusted risk estimates were obtained from original articles, and were used to compare crude summary risk estimates with adjusted summary risk estimates. After log-transforming all estimates, standard errors (SE) were determined by subtracting the lower log-transformed CI boundary from the upper log-transformed CI boundary, and dividing its sum by 3.92. Heterogeneity (Q and I2 statistics) was taken into account by performing a random-effects meta-analysis. Summary risk estimates of random-effects models from maximum likelihood estimations are shown as forest plots for studies with a consistent risk estimate (OR). Potential publication bias was assessed using funnel plot and Egger's test. All analyses were conducted using the metafor-package in R 3.2.1.
Sensitivity analysis. Sensitivity analysis was conducted to verify the robustness of pooled results, and to explore possible causes of heterogeneity. Accordingly, subgroup analyses were performed for study design (case–control study, cohort study), geographic regions (United States of America and Canada, Australia, Europe), trends over time (year of recruitment <1990, 1991-1999, ≥2000), and risk of bias (high risk, low risk).
Results
Literature search. The literature search process (described in Figure 1) identified 41 studies that investigated an association between solarium use and melanoma risk. Six studies were excluded because they did not report any risk estimates for this association (50-55). Duplicate samples were found in four studies, and were therefore redundant (56-59). Finally, our literature search identified two cohort and 29 case–control studies which were eligible for meta-analysis (19-49).
Study characteristics. Characteristics of the included studies are shown in Table I. Most studies were conducted in Europe (64.5%), followed by North America (29.0%), and Australia (6.5%). Samples were mainly recruited before 2000 (80.0%), and differed in age and gender distributions. Overall, included studies comprised 11,706 malignant melanoma cases and 93,236 controls regarding the association between ever exposure to UV radiation from a solarium and melanoma risk (see Table II).
Assessment of study quality, level of evidence and grade of recommendation. The overall quality of studies and the resulting evidence levels were low due to the lack of interventional trials and severe limitations (including unobserved or unrecorded confounding) of many of the observational studies, which might cause a high risk of bias (Tables II and III). It should be emphasized that the results of these cohort and case–control studies represent associations and do not prove causality. Remarkably, in all studies most likely risk of bias resulted in an overestimation of melanoma risk (supplemental file available from the authors upon request). Scores on the modified Newcastle-Ottawa Quality Assessment Scale were on average low, as 67.7% of the 31 included cohort and case–control studies scored less than four stars (supplemental file available from the authors upon request). Assessing all individual studies according to recommendations of the Oxford Centre for Evidence-based Medicine, we defined the association of ever-exposure, first exposure at younger age and high/low exposure to UV radiation from a solarium with melanoma risk as level four of evidence (poor quality cohort and case-control studies) and grade D of recommendation (supplemental file available from the authors upon request). Only a minority of studies reported ORs that were adjusted for the same confounding factors. As many as 35.5% (n=11) of all the included studies did not account for a single confounder. The remaining studies (n=20) adjusted mainly for age (n=15), sex (n=11), and skin color (n=11). Fewer studies adjusted for hair colour (n=10), sun exposure (n=8), sunburns (n=8), family history of melanoma (n=7), naevi (n=7), freckles (n=5) and education (n=5). Moreover, individual confounders were assessed across the included studies differently, and were only partly comparable. Overall, we observed a relatively high heterogeneity across included studies (e.g. ever-exposure: I2=75.98%), and thus performed a random-effects meta-analysis.
Flowchart of the literature search process.
Association of ever-exposure to UV radiation from a solarium with melanoma risk. The summary risk estimate of the random-effects meta-analysis for all studies (cohort and case–control studies combined, as seen in Figures 2, 3 and 4) showed a statistically significant weak association for ever-exposure to UV radiation from a solarium with melanoma risk compared with non-exposure (as seen in Figure 2, overall relative risk=1.19; 95% CI=1.05-1.34; Q(30)=114.33; p<0.001; I2=74.55%). Exclusion of the study by Nielsen et al. (40), which reported a HR instead of an OR, altered results only slightly (Table III; OR=1.19; 95% CI=1.04-1.35; Q(29)=114.33; p<0.001; I2=75.98%). The funnel plot did not show evidence of publication bias (Figure 5, Egger's test; p=0.169).
Sensitivity analyses yielded results inconsistent with our main finding. Subgroup analyses did not show statistically significant associations when separating for geographic region (studies performed in Europe, Figure 2 and Table III; OR=1.10; 95% CI 0.95-1.27; Q(18)=49.39; p<0.001; I2=60.15%), risk of bias (studies with low risk of bias, Figure 3 and Table III; OR=1.15; 95% CI=0.94-1.41; Q(10)=29.63; p=0.001; I2=66.30%), and trends over time (studies conducted after 1990, Figure 4 and Table III; OR=1.09; 95% CI=0.93-1.29; Q(15)=72.97; p<0.001; I2=79.51%). According to the Oxford Centre for Evidence-based Medicine, for the outcome ever-exposure to UV radiation from a solarium, we determined an evidence level of 3a– (systematic review of poor quality cohort and case–control studies) and a grade D of recommendation.
Characteristics of included studies (n=31).
Association of first exposure to UV radiation from a solarium at young age with melanoma risk. Thirteen included studies investigated a possible association between age at first use of a solarium and melanoma risk. However, only four studies reported a risk estimate for the same age threshold (<25 years). For consistency, a meta-analysis was solely performed with these four studies. The summary risk estimate indicated a statistically significant moderate association between first exposure to UV radiation from a solarium before age 25 years and melanoma risk (Table III; OR=1.59; 95% CI=1.38-1.83; Q(3)=1.06; p=0.787; I2=0.00%). According to the Oxford Centre for Evidence-based Medicine, for the outcome first exposure to UV radiation from a solarium at young age” we determined an evidence level of 3a– (systematic review of poor quality cohort and case–control studies) and a grade D of recommendation.
Association of high/low exposure to UV radiation from a solarium with melanoma risk. Several included studies (n=15) determined possible dose–response relationships between exposure to UV radiation from a solarium and melanoma risk. Seven out of these studies used a consistent definition (>10 sessions in lifetime) and were thus appropriate for meta-analysis. The pooled result of this analysis indicated a statistically significant moderate association for high exposure to UV radiation from a solarium with melanoma risk (Table III; OR =1.43; 95% CI=1.17-1.74; Q(6)=19.32; p=0.004; I2=60.87%). However, most of the pooled studies (85.7%) had a high risk of bias. A meta-analysis with the same seven studies was performed for low exposure to UV radiation from a solarium (defined as ≤10 sessions in lifetime) and did not show a statistically significant association (Table III; OR =1.13; 95% CI=0.93-1.38; Q(6)=17.06; p=0.009; I2=58.51%). According to the Oxford Centre for Evidence-based Medicine, for the outcome “high and low exposure to UV radiation from a solarium and melanoma risk” we found evidence level of 3a– (systematic review of poor quality cohort and case–control studies) and a grade D of recommendation.
Risk estimates for included case–control and cohort studies (n=31).
Discussion
Several meta-analyses and reviews have already investigated the relevance of solarium use as a potential melanoma risk factor in recent years. However, most of them have been criticized for limitations, unbalanced view, errors, or incorrectnesses (11, 17). As an example, incorrectnesses in one of the main findings of the study of Boniol et al. (8) forced the authors to publish a correction (16). As Colantonio et al. point out, comparison of five previously published systematic reviews on this topic demonstrates an alarming tendency for copying data without referencing the original article, and without checking for errors (11). As an example, the influential review of the IARC Working group published in 2007 (10) has been criticized for numerous errors in content and typography [e.g. giving wrong numbers for the controls reported 1989 by MacKie et al. (38) and 1981 from Adam et al. (19)], which are also present in two subsequent reviews (11). Furthermore, the numbers of participants from several included studies (31, 43) published in the IARC review could not be derived by us and others (11) from the original articles. Our meta-analysis differs in several points from studies published previously. Firstly, we were able to identify and include some recently published studies that were not included in previous meta-analyses. Secondly, we chose a slightly different approach as compared with most meta-analyses published previously: because included case–control and cohort studies were heterogeneous regarding the adjustment of reported risk estimates (reporting crude/adjusted ORs or HRs; adjusted ORs or HRs were adjusted for greatly varying factors), we decided to use crude risk estimates in a random-effects model for our analysis to improve comparability. Moreover, we investigated the quality of individual studies using a modified Newcastle-Ottawa quality assessment scale and a generally accepted grading system for recommendations in evidence-based medicine.
Forest plot for association of ever-exposure to UV radiation from a solarium with melanoma risk by geographic region. Rounding errors may occur in the forest plot. Weights refer to the 1overall summary risk estimate, 2summary risk estimates of respective subgroups; 3detailed information can be derived from Tables II, III. OR: Odds ratio; CI: confidence interval.
Forest plot for association of ever-exposure to UV radiation from a solarium with melanoma risk by risk of bias. Rounding errors may occur in the forest plot. Weights refer to the 1overall summary risk estimate, 2summary risk estimates of respective subgroups; 3detailed information can be derived from Tables II, III. OR: Odds ratio; CI: confidence interval.
Summary risk estimates from random-effects meta-analyses.
The overall evidence level and quality of studies identified (19-49) was very low due to the lack of interventional trials and because of severe limitations of many of the observational studies. In the meta-analysis of all cohort and case–control studies identified by our literature search, we found a weak association for ever-exposure to UV radiation from a solarium with melanoma risk. Based on 27 studies, the meta-analysis of Boniol et al. (8) reported in 2012 a summary relative risk of 1.20 (95% CI=1.08-1.34) for the association of ever-exposure to UV radiation from sunbeds with melanoma risk (heterogeneity: I2=56%). The authors also estimated that 3,438 (5.4%) of 63,942 new cases of cutaneous malignant melanoma diagnosed each year in the 15 countries that were members of the European Community and the three countries that were part of the European Free Trade Association were related to sunbed use (8). Wehner et al. estimated the population proportional attributable risk of 2.6%-9.4% for melanoma, corresponding to more than 10,000 melanoma cases (12) each year attributable to solarium use in the United States, Europe and Australia. Colantonio et al. reported in their meta-analysis of 31 studies with data available on 14,956 melanoma cases and 233,106 controls an overall OR of 1.16 (95% CI=1.05-1.28) for the association of ever-use of a solarium with melanoma risk (11). While the overall OR of our study (OR=1.19; 95% CI=1.04-1.35, p=0.009) is comparable with the results of Boniol et al. (8) and Colantonio et al. (11), we disagree with their conclusions. In our view, Boniol et al. (8) and Colantonio et al. (11) did not adequately consider the many limitations of the individual studies and the resulting low levels of evidence and grades of recommendation that do not allow postulation of a causal relationship between solarium use and melanoma risk. Moreover, in our opinion, the attempts of Boniol et al. (8) and others (12) to attribute melanoma cases to solarium use are speculative and scientifically not sufficiently supported.
Additionally, our meta-analysis indicated a moderate association of first exposure at younger age and high exposure to UV radiation from a solarium with melanoma risk. However, these results should be interpreted with caution. It should be emphasized that all cohort and case–control studies included in this meta-analysis (19-49) are likely to have overestimated the association of solarium use with melanoma risk in the general population because of many independent reasons, including (i) selection bias (exclusion of individuals with a likely relatively high UV-exposure in the past [e.g. history of any kind of skin cancer or dermatological conditions] in controls, but not in cases), (ii) information bias (e.g. recall bias, the inclusion of non-solarium exposure to artificial UV, e.g. phototherapy), (iii) difficulties in appropriately considering or adjusting for other confounding factors (e.g. solar UV or lifestyle, including smoking), and (iv) the restriction of the analysis to a subgroup of the general population, which may have an increased risk for melanoma (e.g. women).
Forest plot for association of ever-exposure to UV radiation from a solarium with melanoma risk by recruitment period. Rounding errors may occur in the forest plot. Weights refer to the 1overall summary risk estimate, 2summary risk estimates of respective subgroups; 3detailed information can be derived from Tables II, III. OR: Odds ratio; CI: confidence interval.
Like others (37), our study could not confirm the emphasis of the IARC report (10) and of the report by Boniol et al. (8) on an increased melanoma risk with first use of indoor tanning in younger age. It should be mentioned that both the IARC report (10) and the report by Boniol et al. (8) have to be criticized for defining first use in younger age as first use before the age of 36 years, but included studies that consider first use prior to ages 25 to 30 years (23, 42, 55). Moreover, some studies (46, 47) restricted their investigation to melanoma cases diagnosed before the age of 36 years. However, this could have resulted in the exclusion of older cases and controls that may have been exposed at a younger age (37).
Funnel plots for association of ever-exposure to UV radiation from a solarium with melanoma risk. Notes: Funnel plot asymmetry was tested using Egger's test. A: All included studies (n=31), B: studies from North America (n=9), C: studies from Europe (n=20).
The obvious difficulties in considering or adjusting for important confounders are underlined in the Results section. Interestingly, subgroup analyses for studies performed in Europe, studies with low risk of bias, and studies with recruitment 1991-1999 did not show an association of melanoma risk with solarium use.
Concerning our finding of no significant statistical association between ever-exposure to UV radiation from a solarium and melanoma risk in studies performed in Europe (in contrast to studies performed in the United States, several factors are of particular relevance. Firstly, the role of solar UV exposure represents a major confounding factor that is difficult to control or to adjust for, and that may well at least in part explain latitude-dependent variations in melanoma risk. On the other hand, other region-specific factors, which include technical differences in solarium devices, must also be taken into account. Since 2008, solarium devices in Europe and Oceania (Australia and New Zealand) are restricted in intensity to an ultraviolet index of 12 and 36 (which was 60 before 2002), respectively. In contrast, the intensity of a solarium in the United States is unlimited but often a “maximum recommended exposure time” is given.
For many factors that may influence the association of solarium use and melanoma risk, including legal regulations, solarium technology and epidemiology of solarium use, which are subject to frequent change, it is of particular interest to evaluate trends over time. Another interesting observation of our sensitivity and subgroup analyses was the finding that recruitment period had a strong impact on the association of melanoma risk with solarium use. For recruitment before 1991, a higher OR was found as compared with recruitment from 1991-1999 or since 2000. It can be speculated that this observation is due to changes in operation and technical modifications of UV-emitting devices (approximately two decades ago, the solarium industry started to produce devices with higher pressure bulbs emitting larger doses of long-wave UV A). The results of our meta-analysis and previous published studies most likely overestimate the association of melanoma risk with current solarium use as many countries have recently imposed strict regulations on solarium use that, besides other effects, should reduce first use at younger age and high use of a solarium. However, the questions whether stricter regulations of recent years and technical progress have further improved the safety of solarium use are difficult to answer because in many cases, solarium use is not clearly restricted to distinct time periods of interest.
We emphasize that interventional trials are lacking and that the results of the cohort and case–control studies included in this meta-analysis represent associations that do not prove causality. Moreover, both the resulting level of evidence and grade of recommendation of studies investigating the association of melanoma risk with solarium use are weak. According to the Oxford Centre for Evidence-based Medicine, for all outcomes analyzed in our meta-analysis, we found level 3a– of evidence (poor quality cohort and case-control studies) and grade D of recommendation. The poor quality of the cohort and case-control studies included in this meta-analysis is due to severe limitations that include difficulties in appropriately considering and controlling for known confounders (e.g. exposure with solar UV or artificial UV for medical purposes; lifestyle, including smoking).
In summary, our review has highlighted the poor quality of the evidence available at present on this topic. We conclude that (i) results of our and previously published meta-analyses most likely overestimated the association of melanoma risk with solarium use, (ii) both the level of evidence and grade of recommendation of studies published previously investigating the association of melanoma risk with solarium use are weak, and therefore (iii) present scientific knowledge does not support the hypothesis of an increased melanoma risk due to solarium use, and questions studies that try to attribute melanoma cases to indoor tanning, and does not support initiatives that aim to ban responsible/moderate solarium use for tanning purposes.
Footnotes
↵* These Authors contributed equally to this study.
Author's Disclosures
BB, JS, IH, CS, SR, SW, SP, JOM, WM have nothing to disclose. Saarland University (TV, JR) received funding from Jörg Wolff Foundation and Immundiagnostik AG. JR received speakers honoraria from Wörwag and Dermapharm.
- Received October 31, 2017.
- Revision received December 13, 2017.
- Accepted December 29, 2017.
- Copyright© 2018, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
References
In this issue
