Melanocortin 1 receptor and risk of cutaneous melanoma: A meta-analysis and estimates of population burden



Polymorphisms in the melanocortin 1 receptor (MC1R) gene have been associated with increased risks of melanoma, but different approaches to study design, analysis, and reporting have hindered comparisons of findings. We aimed to harmonize the published data by conducting a systematic review and meta-analysis of MC1R variants and thereby estimate relative risks and population attributable fractions (PAFs). We identified 20 analytic studies reporting on 25 populations, which presented quantitative data on melanoma risks associated with any of nine MC1R variants. We separately pooled estimates of risk per person and risk per chromosome using a random effects model. Red hair color (RHC) variants had the highest risk of melanoma [summary odds ratios (OR) 2.44, 95% confidence interval (CI) 1.72–3.45, PAF 16.8% CI 0.119–0.202], but non-RHC variants were also associated with increased risk (summary OR 1.29, 95% CI 1.10–1.51, PAF 7.4% CI 0.030–0.112). The summary risk of melanoma associated with individual variants ranged from OR 2.40 for R142H to 1.18 for V60L, although significant heterogeneity was evident for most variants. PAFs ranged from 0.55% for I155T to 6.28% for R151C. Our findings suggest the nine most common MC1R variants make a sizeable contribution to the burden of melanoma. Melanoma research would be greatly assisted by standardized classifications for MC1R variants and consistent reporting conventions. More compatible and comparable research would allow for more powerful data that could be clinically applied to predict melanoma risk.


Melanoma is the most lethal type of skin cancer, and incidence continues to increase in fair-skinned populations worldwide.1 Considerable research has been conducted on the etiology of melanoma in recent decades; ultraviolet radiation exposure has been identified as the principal environmental cause, but host factors, primarily genetic, also play a major role.2 In particular, variation in several pigmentation genes has been significantly associated with melanoma susceptibility.3–8 Of these, the gene encoding the melanocortin 1 receptor (MC1R), the receptor for α-melanocyte stimulating hormone (α-MSH), has been studied most extensively. MC1R contributes to pigmentation by regulating the relative concentrations of eumelanin (brown/black pigment) and pheomelanin (red/yellow pigment).9 When α-MSH binds and stimulates the melanocortin receptor on the surface of the melanocyte, cAMP production increases, stimulating eumelanin production.9

MC1R is highly polymorphic within Caucasian populations.10 A recent review of MC1R frequencies documented 57 nonsynonymous and 25 synonymous polymorphisms.11 These polymorphisms have varying effects on the receptor's function, its ability to modify melanogenesis and thus, the eumelanin/pheomelanin ratio.12, 13 A nonfunctional melanocortin receptor will lead to a predominant pheomelanin concentration and what is known as the red hair color (RHC) phenotype, recognized constitutively by the co-occurrence of fair skin, red hair, freckles and poor tanning ability.14–16MC1R variants associated with RHC as a result of nonfunctional receptors, which appear to act in a dominant manner, are labeled R alleles. Variants giving rise to receptors without a total loss of function and a weak or nonexistent association with RHC are called nonred hair color (NRHC) variants, labeled r alleles.15, 17–19

Given the long-standing observation that melanoma risk is highest amongst people with sun-sensitive phenotypes, numerous studies have been conducted to quantify the risk of melanoma associated with pigmentation gene polymorphisms, particularly the RHC and NRHC variants of MC1R.3–8 However, there has been great diversity in both the approach to analyses and in the reporting of findings, presenting challenges for deriving summary estimates of melanoma risk. Variations between studies include the way in which allele frequency has been reported (per person or per chromosome), the definition of the reference groups used to estimate relative risks (e.g., some have compared people with a specific MC1R variant to those without that particular MC1R variant, whereas others have compared with those having no known MC1R variants at all) as well as differences in RHC/NRHC categorization. Taken together, these differences likely contribute to the considerable heterogeneity in risk estimates reported across studies.

Here, we present the findings of a systematic review and meta-analysis of the association between MC1R genotypes and melanoma risk. Our aims were to document the various approaches taken to quantify risk, to harmonize as much as possible the published data, and then derive summary estimates of relative risk of cutaneous melanoma associated with the common variants of MC1R.

Material and Methods

Eligibility criteria

Included in the meta-analysis were case-control and nested case-control studies, which presented quantitative data on the association between at least one of nine prespecified MC1R variants, or RHC/NHRC variants, and melanoma. To be eligible, studies had to report either allele frequencies for cases and controls and/or odds ratios (OR) with 95% confidence intervals (CIs). We used only a single dataset for each eligible study; when a study sample was reported in more than one article, we abstracted the data from the most recent or the most complete article.

Literature search

An electronic literature search identified eligible studies published until October 2009 in the Medline database (U.S. National Library of Medicine, Bethesda, MD, 1950–present) using the search interface, PubMed and the ISI Science Citation Index database using the Web of Science® search interface. References from retrieved articles were also hand searched for additional articles. The following MeSH terms or text words were used for electronic searches: “melanoma,” “receptor, melanocortin, type 1,” “pigmentation gene,” “etiological,” “etiology,” “group,” “risk,” “cohort studies” and “case-control studies” (Box 1). The search included both UK and US spellings and was neither limited to adult or human populations, as these limits eliminated relevant articles, nor was it limited to studies published in English.

The initial search identified 173 potentially relevant studies. After reviewing the titles of these studies, we found 105 articles to be animal or molecular studies, or studies of inappropriately selected populations. We then read the abstracts of the 68 remaining studies and eliminated 19 more articles. Of the remaining 49 articles, 27 described original research, whereas 22 were review articles or meta-analyses; we then hand-searched the reference lists of these 22 articles and retrieved one additional original research article.

Scheme 1.

Search strategy to identify observational studies on the association of MC1R variants and melanoma.

After a full text review of the 28 articles, we discarded five as they reported on the same populations as more recent or more complete articles and three due to insufficient and/or irrelevant quantitative data. Thus, we abstracted data from 20 articles3–7, 20–34 reporting on 25 different populations; all case-control studies, one nested.25 A flow chart of selected studies can be found in Supporting Information Figure 1.

Figure 1.

Forest plot of the association between RHC variants and melanoma using a random effects model. Each line represents an individual study result with the width of the horizontal line indicating 95% CI, the position of the box representing the point estimate, and the size of the box being proportional to the weight of the study. [Color figure can be viewed in the online issue, which is available at]

Data abstraction

Two independent reviewers (P.F.W. and C.M.O.) completed an abstraction form summarizing study design, study population and data on specific variants of interest, RHC and NRHC for each study population. We treated studies reporting on multiple populations with separate data sets independently in the meta-analysis and abstracted their data as such. Two studies reported on two populations,7, 20 and one study on three populations.3 One study reported data for specific variants on a combination of two populations, although reporting RHC and NRHC data for the individual populations; the data were abstracted so as not to give greater weight to the individual populations for either specific variants or RHC and NRHC.21 The following information was recorded for each study: study design, geographic location, years of data collection, number and source of cases and controls, age and family history of melanoma of study populations, Hardy–Weinberg Equilibrium (HWE) testing, and laboratory methods for detecting MC1R variants.

Individual studies presented the number of variants in study populations in several different formats: number of variants per number of individuals, number of variants per number of chromosomes and more specifically as the number of heterozygote and homozygote individuals in case and control populations. We deemed all of these forms acceptable and noted them during the abstraction process. When the number of variants was presented in terms of numbers of heterozygote and homozygote individuals, we were able to convert the data into variant frequencies among people and among chromosomes (per person and per chromosome, respectively), using the equations described by Cordell and Clayton.35

For the nine variants of interest, RHC and NRHC, we recorded the frequencies, ORs and 95% CIs, reference groups and all variables adjusted for in final models. If several risk estimates were given, those adjusted for the greatest number of potential confounders were recorded. We also recorded specific definitions for RHC and NRHC, including which MC1R variants were classified in each category.

We chose the following variants for analysis: V60L, D84E, V92M, R142H, R151C, I155T, R160W, R163Q and D294H, as these were previously determined by a large population-based study as having a minor allele frequency of greater than 1% in at least one population studied.36 RHC vs. NRHC categorizations were those used in the same population-based study, made on the basis of consistent association with the RHC or NRHC phenotype, with RHC variants designated as D84E, R151C, R160W and D294H and NRHC as V60L, V92M, R142H, I155T and R163Q.36 These designations have been inconsistent in the literature, but reflect an association between genotype and phenotype, and we have employed them for ease of comparability with other research.

We did not assess the methodologic quality of the primary studies because the use of quality scoring in meta-analyses of observational studies is controversial.37 Therefore, we did not exclude studies on the basis of quality score, but instead performed subgroup and sensitivity analyses according to study features that could potentially affect the strength of the associations.

Statistical analysis

To pool risk estimates for the presence of a specific variant, a weighted average of the log OR was estimated, taking into account the random effects using the DerSimonian and Laird method.38 We separately pooled estimates of risk per person and risk per chromosome. Statistical heterogeneity among studies was evaluated using the Cochrane Q test and I2 statistics. The Cochrane Q test, calculated as the weighted sum of squared differences between individual study effects and the pooled effect across studies, is widely known to have too much power to detect clinically unimportant heterogeneity if the number of studies is large.39 The I2 statistic describes the percentage of variation across studies that is due to heterogeneity rather than chance40 and does not inherently depend on the number of studies considered (I2 = 100% × (Q − df)/Q).

We performed separate subgroup analyses by sources of cases and controls (population vs. clinic/hospital vs. other sources), the reference groups used to calculate ORs, geographic location, HWE testing and study size (<150 cases vs. ≥150 cases). Finally, we conducted sensitivity analyses, omitting each study in turn to determine whether the results could have been influenced excessively by a single study. We evaluated publication bias by assessing funnel-plot asymmetry using both the Begg and Mazumdar41 and Egger et al. methods.42

Population attributable fractions (PAFs) for the presence of a specific variant, or for any RHC/NRHC-categorized variant, were estimated using the adjusted relative risk (RR) derived from meta-analysis and the method of Bruzzi et al.43 which accounts for possible confounding and effect modification. CIs for the PAFs were derived using the substitution method described by Daly.44 We estimated the prevalence of each MC1R variant by calculating the mean prevalence, weighted by the size of each case group.

Because of heterogeneity in the prevalence of variants across studies from different locations, we decided a priori to calculate PAFs for different geographic locations [North America, Northern Europe, Central Europe, Mediterranean Europe, Australia and Other (Israel)].

All analyses were conducted using Stata 10 (College Station, TX).


Descriptive characteristics of the 25 populations included in the meta-analysis are presented in Table 1. We described sources of cases and controls as “Clinic/hospital,” “Population,” “Not specified” or “Other” (e.g., family members, blood donors, hospital volunteers and electoral roll); case groups from five of the 25 populations were population-based (Table 1).

Table 1. Characteristics of the 25 populations from the 20 case-control studies included in the meta-analysis of MC1R variants and melanoma risk
inline image

The pooled RRs for each individual variant as well as groups of RHC and NRHC variants are shown in Table 2 along with heterogeneity calculations. We did not have sufficient data to pool RHC and NRHC values per chromosome. The estimated summary relative risks for RHC and NHRC variants were 2.44 (95% CI 1.72–3.45) and 1.29 (95% CI 1.10–1.51), respectively, although both estimates were associated with evidence of substantial heterogeneity across studies. R151C, which has been classified consistently as an RHC allele, had a summary relative risk of 1.93 per person (95% CI 1.54–2.41), again with significant heterogeneity across studies (I2 value of 72.4%).

Table 2. Allele frequency in control and case populations from data presented per person, the number of studies reporting on a specific variant and those studies' case and control numbers, odds ratios (OR) with 95% confidence intervals (CIs) for the association between MC1R variants and melanoma, and heterogeneity estimates
inline image

Figures 1 and 2 depict the forest plots of RHC and NRHC variants, respectively, and include the geographic location of each study population. The RHC forest plot shows that all studies reported an increased risk of melanoma associated with carriers of RHC variants except for the Icelandic population.3 When we excluded this population from the meta-analysis, the pooled OR changed only modestly from 2.44 to 2.41 but with narrower CIs. The forest plot for NRHC displays a wider range of relative risks with eight studies reporting positive associations with melanoma and three studies reporting negative associations. Forest plots for each specific variant can be found in Supporting Information Figures 2–10.

Figure 2.

Forest plot of the association between NRHC variants and melanoma using a random effects model. Each line represents an individual study result with the width of the horizontal line indicating 95% CI, the position of the box representing the point estimate, and the size of the box being proportional to the weight of the study. [Color figure can be viewed in the online issue, which is available at]

Subgroup analysis

We repeated the above analyses within subgroups of study design (source of cases/controls), study size, geographic location, choice of reference category (i.e., wild-type MC1Rvs. noncarrier of specific variant), and whether HWE was tested to determine whether these factors altered the summary estimates (Supporting Information Tables 1–11). We observed significant heterogeneity between study designs and no one particular source of cases or controls yielded consistently higher or lower risk estimates. The pooled effect estimates were generally higher for the group of studies that included <150 cases; however, excluding these studies did not result in significant changes to the pooled ORs for all studies (Supporting Information Tables 1–11).

We observed consistently higher estimates of melanoma risk when relative risks were calculated using reference groups comprised of people with only wild type MC1R when compared with reference groups comprised of people who were not carriers of the specific MC1R variant under consideration (but who may have been carriers of other MC1R variants). Moreover, the number of studies using wild type MC1R for the reference group was much smaller (at most four studies for any variant) than the number of studies using alternatives (approxiamtely 12 studies analyzed for each variant). No consistent patterns were observed by geographic location, and studies that did not report HWE testing did not yield consistently higher or lower estimates than those that did. Only one study reported significant departure from HWE6 and only for the D84E variant; excluding this study from the meta-analysis for this variant did not significantly affect the pooled estimate.

Sensitivity analysis and publication bias

Using the Begg rank correlation method of detecting publication bias, there was evidence of bias for data published on the risk of melanoma for persons with the variant I55T. The Egger linear regression method of testing publication bias confirmed this finding and also suggested publication bias for V60L, V92M, R142H and R151C risk per chromosome, and R163Q and RHC risk per person. Sensitivity analysis conducted by excluding one study at a time resulted in summary relative risks for RHC ranging from 2.32 (95% CI 1.63–3.32), when the Stratigos et al. study26 was excluded, to 2.49 (95% CI 1.66–3.74), when the Gubdjartsson et al.3 study was excluded. The summary relative risks for NRHC ranged from 1.25 (95% CI 1.07–1.47), when the Stratigos et al. study26 was excluded, to 1.34 (95% CI 1.15–1.55), when the Brudnik et al.4 study was excluded.

Population attributable fractions

We calculated the PAF for each variant, RHC and NRHC, using the summary relative risks and the weighted average of the prevalence estimates for all studies (Table 3). The PAF values for RHC and NRHC were 16.8% (95% CI 11.9–20.2%) and 7.4% (95% CI 3.0–11.2%), respectively. Table 4 presents PAF values for RHC and NRHC stratified by geographic location. PAFs stratified by geographic location for each specific variant are given in Supporting Information Tables 13a–13i.

Table 3. Estimates of the population attributable fraction (PAF) of melanoma associated with MC1R variants
inline image
Table 4. Estimates of the population attributable fraction (PAF) of melanoma associated with RHC and NRHC variants stratified by geographic location
inline image


We have systematically reviewed the literature reporting on the relationship between melanoma and MC1R variants and conducted a meta-analysis using data from 8,636 cases and 50,661 controls. We found the highest summary risk of melanoma per person associated with RHC variants with a greater than twofold increased risk; NRHC variants were associated with a modest increased risk. All individual variants were associated with an increased risk of melanoma with summary estimates ranging from 1.2 to 2.4. The PAFs for individual variants ranged from 0.55% for I155T to 6.28% for R151C.

A key finding of the systematic review was the substantial variation in study design, analytic approaches and reporting of relative risks in the published literature, considerably influencing the interpretation of findings. Only five of the 25 populations included cases recruited from population sources. Selection bias due to recruitment from dermatology clinics or hospitals in the nonpopulation-based case–control studies would result in an attenuated effect, if controls recruited in this way were more likely to be MC1R variant carriers than those recruited randomly from a population-based source.

Inconsistent definitions for RHC and NRHC variants restricted our ability to compare and combine study findings. Thus, although R151C, R160W and D294H were universally classified as RHC variants and V60L, V92M and R163Q were classified as NRHC variants, less common variants such as I155T and R142H were either not investigated at all or classified differently across studies. For example, R142H was classified as an RHC allele in four studies4, 20, 26, 27 and as NRHC in two21, 28, whereas I155T was classified as RHC in one study27 and NRHC in four3, 4, 20, 21. We found that NRHC was a less consistent grouping than RHC, with some studies not providing a definition of the variants composing their NRHC classification at all, and some studies including all nonsynonymous variants in this group. Including all nonsynonymous variants in NRHC complicates comparisons between studies as these groups will have both different and a greater number of variants.

Another source of heterogeneity between studies was the choice of study comparator group (controls). Although it is customary in genetic studies to compare the risk of disease among people with a certain genetic variant to those without that variant, MC1R polymorphisms are so common that this practice will inevitably lead to biased risk estimates. Absence of the “variant-of-interest” is not synonymous with “wild-type” sequence; it is just as likely for an individual to have another MC1R variant than no MC1R variant becasue 50% of the Caucasian population has at least one polymorphism in this gene.10 Therefore, individuals lacking a specific variant are not necessarily at baseline risk, and may potentially have an elevated risk of disease compared with individuals carrying only “wild type” genes. Such bias may underestimate the true risk of disease.

Beyond data collection and calculation, we also noted considerable variation in terms of reporting. Some studies reported risk of melanoma as risk per chromosome, whereas others described risk per person. Studies that reported risks in terms of chromosomes did not distinguish between heterozygote and homozygote individuals and were deemed incompatible with studies that reported risk in terms of individual people. Although we were able to calculate separate estimates determined by the manner of reporting, a stronger meta-analysis could be performed if reporting were standardized.

The issue of different approaches to reporting MC1R risks has been noted previously.45 Our study differs from this earlier review by estimating separately risk per chromosome and per individual. When comparing our risk estimates per chromosome and the reported risks in the previous meta-analysis, we confirmed the earlier finding of increased melanoma risk for carriers of the uncommon variants D84E and R142H; indeed these variants conferred the highest summary risks in our meta-analysis. Our study shares some features with the recent meta-analysis by Kanetsky et al.,46 which also found increased melanoma risk in “high risk” variant carriers with certain phenotypes. However, that analysis found no association between melanoma and carriage of only one “low risk variant.” Although their definition of “high risk” is the same as our RHC classification (D84E, R151C, R160W and D294H) and the reference group used to calculate risks were individuals with no MC1R variant, the “low risk variant” category was composed of all nonsynonymous variants as opposed to specifying NRHC variants. The latter analyses included only studies that reported risks stratified by phenotypic characteristics; six of the seven studies included in the Kanetsky meta-analysis were included in the current meta-analysis. Kanetsky et al. reported that the risk of melanoma associated with MC1R variant genotypes was higher in individuals with protective phenotypes and limited sun exposure. They suggested that MC1R genotype status might provide additional information about melanoma risk in people who would not be identified as high risk based on their phenotypes or exposures alone. Thus, the possibility arises that MC1R variants may mediate their effects through biological pathways that are independent of pigmentation. Recent studies implicate MC1R in a number of key regulatory pathways involved in cell cycle control47 and apoptosis.48

Our analysis suggests that some publication bias may also have contributed to the heterogeneity among studies, at least for some variants. The greater extent of bias detected by the Egger and coworkers method49 can be explained by its greater sensitivity and statistical power when compared with the Begg method. However, both methods can give false positives when the number of studies is small. As we never evaluated more than 20 studies at a time, we cannot rule out that the distribution of published findings was simply due to chance and not systematic publication bias.49

In summary, the published data suggest that all common MC1R variants are associated with modestly increased relative risks of melanoma and that some of the less common RHC variants confer reasonably high relative risks on carriers. Importantly, the modestly increased risk of melanoma associated with NRHC variants suggests that MC1R may be involved in melanoma development through pathways that are independent of the classical RHC phenotype. Delineating such pathways should be the focus of further research. Our attempt to harmonize the extant literature is a necessary first step for developing improved risk prediction models for melanoma. Finally, given the heterogeneity of existing reports of MC1R with respect to melanoma, we encourage the standardization of RHC/NRHC classifications and suggest the continued use of those presented in this review. Additionally, we suggest adopting consistent reporting conventions for MC1R allele frequencies in terms of variants per chromosome for more compatible future analyses and comparisons.


This research was supported by a grant from the Xstrata Community Partnership Program Queensland.

The researchers are independent of the funding source. N.K.H. is supported by a Research Fellowship from the National Health and Medical Research Council of Australia, and D.C.W. is supported by a Future Fellowship from the Australian Research Council.