SEARCH

SEARCH BY CITATION

Keywords:

  • melanoma;
  • eye colour;
  • hair colour;
  • skin phototype;
  • freckling

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Epidemiologic research has demonstrated convincingly that certain pigmentary characteristics are associated with increased relative risks of melanoma; however there has been no comprehensive review to rank these characteristics in order of their importance on a population level. We conducted a systematic review of the literature and meta-analysis to quantify the contribution of pigmentary characteristics to melanoma, estimated by the population-attributable fraction (PAF). Eligible studies were those that permitted quantitative assessment of the association between histologically confirmed melanoma and hair colour, eye colour, skin phototype and presence of freckling; we identified 66 such studies using citation databases, followed by manual review of retrieved references. We calculated summary relative risks using weighted averages of the log RR, taking into account random effects, and used these to estimate the PAF. The pooled RRs for pigmentary characteristics were: 2.64 for red/red-blond, 2.0 for blond and 1.46 for light brown hair colour (vs. dark); 1.57 for blue/blue-grey and 1.51 for green/grey/hazel eye colour (vs. dark); 2.27, 1.99 and 1.35 for skin phototypes I, II and III respectively (vs. IV); and 1.99 for presence of freckling. The highest PAFs were observed for skin phototypes 1/II (0.27), presence of freckling (0.23), and blond hair colour (0.23). For eye colour, the PAF for blue/blue-grey eye colour was higher than for green/grey/hazel eye colour (0.18 vs. 0.13). The PAF of melanoma associated with red hair colour was 0.10. These estimates of melanoma burden attributable to pigmentary characteristics provide a basis for designing prevention strategies for melanoma.

Among Caucasians, a large body of evidence suggests that pigmentary characteristics are important determinants of melanoma susceptibility; however, to date, there has been no comprehensive review to rank these characteristics in order of their importance on a population level. The identification of individuals at high risk of developing melanoma, who might be targeted for prevention and screening efforts, requires an understanding not only of the magnitude of the risk associated with each factor, but also their contribution to the burden of disease at the population level. The population-attributable fraction (PAF) is widely used to quantify the public health impact of putative causal factors, since it considers both the strength of association between risk factor and disease, as well as the prevalence of the factor in the community.1

We have previously evaluated the PAF of melanoma associated with common and atypical nevi2 and family history of melanoma.3 The aim of this work was to evaluate systematically the most recent epidemiological evidence describing the relationship between melanoma and pigmentary characteristics and to use these data to estimate the fraction of melanomas attributable to these phenotypic characteristics.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

A systematic review and meta-analysis was performed in accordance with MOOSE guidelines for reviews of observational studies.4

Eligibility criteria

We included observational studies of all designs in the meta-analysis provided that they permitted quantitative assessment of the association between histologically confirmed melanoma and eye colour, hair colour, skin phototype, skin colour and/or freckling. We included studies reporting various measures of relative risk (RR) since melanoma is a rare disease, and in such instances, odds ratios (ORs) and standardised incidence ratios (SIRs) provide a valid estimate of the RR.

Literature search

Eligible studies up to June 2008 were identified by searching the following databases and by hand-searching the reference lists of the retrieved articles.

  • Medline 1950 (U.S. National Library of Medicine, Bethesda, MD), using PubMed software as the search interface

  • Embase 1966 (Elsevier Science, Amsterdam, Holland), using the Embase search interface

  • Conference Papers Index 1982 (CSA, Bethesda, MD), using the CSA Illumina search interface

  • ISI Science Citation Index, using the ISI Web of Science® search interface

For computer searches, we used the following MeSH terms or text words (using both UK and US spellings): ‘Pigmentation’[Mesh], ‘Eye Colour’[Mesh], ‘Hair Colour’[Mesh], ‘skin colour’, ‘skin phototype’, ‘phenotype’, ‘phenotypic’, ‘host’, ‘hair colour’, ‘eye colour’, ‘host-related’, ‘constitutional’, ‘pigmentary’, ‘freckl*’, ‘ephelid*’, ‘risk’, ‘etiological’, ‘aetiology’, ‘cohort studies’, ‘case-control studies’. Studies that had been commonly cited in the literature were also included as citation search terms in the ISI Science Citation Index (1990–present) to identify subsequent studies that had referenced them. Only studies of adult populations (>18 years) were included. The search was not limited to studies published in English. We read the abstracts of all identified studies to exclude those that were clearly not relevant. The full texts of the remaining articles were read to determine if they met the study inclusion criteria. Where multiple reports from one study were found, the most recent or most complete publication was used.

Data extraction

An abstraction form summarising study design, study population and relevant raw and adjusted data was completed for each article by 2 independent reviewers (CO, HC); inconsistencies were resolved by consensus. The following information was recorded for each study: study design, location, calendar years of data collection (case–control studies), number of cases and controls or person-years duration of follow-up (cohort studies), age range of study population, variables for which statistical adjustment was done, point estimates (RR, OR or SIR) and 95% CIs. Definitions of the various phenotypic variables were also recorded together with information on data collection (e.g., self-reported vs. observer verified). Where several risk estimates were presented, we abstracted those adjusted for the greatest number of potential confounders. Studies that reported results separately by gender, body site or histological subtype with no combined data were treated as independent datasets in the meta-analysis.

We did not assess the methodological quality of the primary studies and hence did not exclude studies on the basis of quality score,4 but instead performed subgroup and sensitivity analyses according to study features that could potentially affect the strength of the association.

Statistical analysis

To pool RR estimates for pigmentary variables, a weighted average of the log RR was estimated, taking into account the random effects using the method of DerSimonian and Laird.5 Statistical heterogeneity among studies was evaluated using the Cochrane Q test and I2 statistics. The Cochrane Q test, which is calculated as the weighted sum of squared differences between individual study effects and the pooled effect across studies, is widely regarded as having too much power to detect clinically unimportant heterogeneity when the number of studies is large.6 The I2 statistic describes the percentage of variation across studies that is due to heterogeneity rather than chance,7 and does not inherently depend upon the number of studies considered (I2 = 100% × (Qdf)/Q). We also conducted separate analyses by study design, geographic location, year of publication (before and after 2000), whether the characteristic was self-reported or observed, and by confounders controlled for in the analyses (other pigmentary characteristics, nevi). 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.8, 9

To estimate the PAF for each characteristic using the adjusted RR derived from meta-analysis, we used the method of Bruzzi et al.,10 which accounts for possible confounding and effect modification. CIs for the PAFs were derived using the substitution method described by Daly.11 We estimated the prevalence of each pigmentary risk factor by calculating an average of the prevalences reported in study cases from population-based studies, weighted by the size of the case group. Where there was heterogeneity in prevalence of a pigmentary trait across studies we conducted sensitivity analyses to evaluate the impact of the range of prevalence on the PAF estimates. All analyses were conducted using Stata 10 (College Station, TX, USA).

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The primary computerised literature search identified 124 potentially eligible studies. After review of the study abstracts, we retrieved 92 articles for further assessment, of which 3 reports of cohort studies,12–14 one of a nested case–control study,15 and 62 from case–control studies (total = 66) met the eligibility criteria (Table 1). Of the eligible case–control studies, 20 were population-based16–35 and 42 were clinic/hospital-based.36–77 The remaining 26 studies were excluded because they either did not present enough data to compute effect estimates of the associations (n = 9), were not independent of other included studies (n = 11), or they represented an ineligible study design (e.g., case-series, non-population-based cohort study) (n = 6).

Table 1. Characteristics of the 66 studies included in the meta-analysis of pigmentary characteristics and risk of melanoma
inline image

Hair colour

A total of 56 studies presented data on the association between hair colour and melanoma: 3 cohort,12–14 one nested case-control,15 28 population-based case–control16–20, 22–34 and 24 clinic/hospital-based case–control studies.36–41, 44–58, 60–62, 64–68, 71–75 Forty-nine studies presented effect estimates for risk of melanoma associated with red hair,12–20, 22, 24–34, 37–41, 44–46, 48, 50–52, 54–56, 58, 60–62, 64–68, 71, 73–75 24 for blond hair,12–14, 18–20, 22, 24, 29, 31, 32, 34, 37, 44, 47, 53, 54, 58, 60, 62, 66–68, 75 and 22 for light brown hair.14, 16–19, 26, 29, 30, 32–34, 39, 41, 45–47, 53, 55, 56, 60, 68, 74

Using a random effects model, the pooled RR for risk of melanoma associated with red hair compared to dark hair was 2.64 (95% CI: 2.25–3.10) for all studies and was slightly higher for population based studies than clinic/hospital-based studies (3.01 vs. 2.29) (Table 2). There was evidence of significant heterogeneity overall (p < 0.001), and for case–control studies (p < 0.001) but not cohort studies (p = 0.151). When stratified by geographic location, the highest pooled RR was observed for the 7 studies conducted in Central Europe (pooled RR: 3.08, 95% CI: 1.92–4.94) and the lowest for those conducted in Australia (pooled RR: 2.10, 95% CI: 1.36–3.22). The pooled RR varied by definition of red hair used: RR 3.11 (95% CI: 2.42–4.01) for ‘red’; 2.05 (95% CI: 1.77–2.37) for ‘red or red/blond’.

Table 2. Meta-analysis results using a random effects model: Risk of melanoma associated with hair colour
inline image

The pooled RR for risk of melanoma associated with blond hair compared with dark hair was 2.00 (95% CI: 1.47–2.73), again with significant heterogeneity overall (p > 0.001), and for case–control studies (p > 0.001) but not cohort studies (p = 0.435) (Table 2). The pooled RR was higher for population-based studies than for clinic/hospital-based studies (2.77 vs. 1.37). The pooled RR for risk of melanoma associated with light brown hair colour compared with dark hair was 1.46 (95% CI: 1.26–1.68) with evidence of significant heterogeneity (p > 0.001). Again the pooled estimate was slightly higher for population-based studies compared with clinic/hospital-based studies (1.58 vs. 1.28), and no heterogeneity was observed for the population-based studies (p = 0.146).

For all hair colour variables, the pooled RR was slightly attenuated when restricted to studies that had adjusted for eye colour and/or skin phototype, and the pooled RR was slightly higher when hair colour was self-reported rather than observer recorded. Restricting the analyses to studies that had adjusted for one or more nevi variables resulted in lower pooled RRs for red and blond hair, but not light brown hair. Sensitivity analyses excluding one study at a time revealed that no one study was unduly influencing the pooled RRs.

Eye colour

A total of 44 studies presented data on the association between eye colour and melanoma: 2 cohort,13, 14 15 population-based case–control16, 17, 20, 22–26, 28–34 and 27 clinic/hospital-based case–control studies.36, 38–42, 44, 46–55, 57, 58, 60, 61, 64, 68, 71–74

For the association between blue or blue/grey eye colour and melanoma, 31 studies contributed to the meta- analysis.13, 16, 17, 20, 22, 24–26, 28–34, 38, 39, 44, 46, 47, 50–52, 55, 58, 60, 64, 68, 71, 72, 74 For the study by Zanetti et al.44 we combined the categories ‘light blue’ and ‘blue-grey’. The pooled RR for the risk of melanoma associated with blue/blue-grey eye colour (compared with dark) was 1.57 (95% CI: 1.39–1.78) with evidence of significant heterogeneity (p < 0.001) (Table 3). The pooled RR was slightly higher for clinic/hospital-based studies than for population-based studies (1.67 vs. 1.55).

Table 3. Meta-analysis results using a random effects model: Risk of melanoma associated with eye colour
inline image

Twenty-five studies contributed to the meta-analysis of green/grey/hazel eye colour.13, 14, 16, 17, 22, 24, 26, 29–34, 39, 44, 46, 47, 55, 58, 60, 61, 64, 68, 72, 74 We combined eye-colour categories for 3 studies as follows: Wolf et al.,45 ‘green’ and ‘grey’; Grulich et al.,60 ‘green’ and ‘hazel’; and Holman et al.,34 ‘green’, ‘grey’ and ‘hazel’. The pooled RR for risk of melanoma associated with green/grey/hazel eye colour (compared with dark) was 1.51 (95%CI: 1.28–1.79) again with evidence of significant heterogeneity (p < 0.001) (Table 3). The pooled RR was lower and non-significant for the 2 cohort studies (1.23, 95% CI: 0.87–1.74) with no evidence of heterogeneity (p = 0.73).

For both eye colour variables, the pooled RR was slightly attenuated when restricted to studies that had adjusted for hair colour and/or skin type/skin colour. Sensitivity analyses excluding one study at a time revealed that no one study was unduly influencing the pooled RRs.

Skin phototype

A total of 42 studies presented data on skin phototype and melanoma risk, and these were all case–control studies; 14 were population-based16, 18–24, 27, 30–34 and 28 were clinic/hospital-based.36–42, 44, 45, 47, 48, 50–54, 56, 58–62, 64, 65, 68–70, 76 Eighteen studies presented data on 4 levels of skin phototype (I, II, III, IV),16, 18–20, 22, 24, 27, 31, 32, 34, 36, 45, 54, 58–60, 64, 69 and a further 23 studies presented data on phototypes I/II vs. III/IV.21, 23, 30, 33, 37–42, 44, 47, 48, 50–53, 56, 61, 62, 68, 70, 76 Eight studies presented data in 3 categories of skin phototype33, 39, 47, 53, 61, 65, 68, 70; we were able to collapse 2 of these categories for all of these studies except one65 and include them in the meta-analysis of skin phototype I/II vs. III/IV.

Compared with skin phototype IV, the pooled RR for risk of melanoma was 2.27 (95% CI: 1.77–2.92) for phototype I, 1.99 (95% CI: 1.62–2.45) for phototype II and 1.35 (95% CI: 11.12–1.63) for phototype III (Table 4), with evidence of significant heterogeneity (p < 0.001). The pooled RRs for studies that adjusted for hair colour and/or eye colour were lower than for studies that did not adjust for these pigmentary variables. There was a modest degree of variation by geographic location. For the 23 studies that compared skin phototypes I/II to III/IV, the pooled OR for risk of melanoma associated with phototypes I/II was 2.28 (95% CI: 1.90–2.73), again with evidence of significant heterogeneity (p < 0.001) (Table 5). The pooled RR was similar for population-based and clinic/hospital-based case–control studies, although heterogeneity was restricted to the clinic/hospital-based studies. Pooled estimates were higher when skin phototype was observed rather than self-reported and again was lower for studies that had adjusted for eye and/or colour or hair colour.

Table 4. Meta-analysis results using a random effects model: Risk of melanoma associated with skin phototype
inline image
Table 5. Meta-analysis results using a random effects model: Risk of melanoma associated with skin phototype I or II vs III or IV
inline image

Skin colour

A total of 30 studies presented data on skin colour and melanoma risk: one cohort,14 9 population-based case–control studies16, 23, 24, 26, 28, 29, 32, 33, 35 and 20 clinic/hospital-based case–control studies.36, 38, 39, 41, 43, 46, 49, 50, 53–58, 61, 63, 67, 68, 72, 75 The definitions of skin colour varied widely across studies, as did the number of categories of skin colour (19 studies presented data in 2 categories, 10 in 3 categories and 1 study presented 4 categories of skin colour). Because of this heterogeneity in data presentation, we were unable to pool the study-specific estimates for meta-analyses of skin colour and melanoma risk.

Freckling

Thirty-nine studies presented data on presence of freckling and risk of melanoma, all case–control studies: 15 population-based16–21, 23–29, 32, 33 and 24 clinic/hospital-based.38, 41, 43, 44, 46, 49–56, 58, 60, 61, 63, 64, 66, 68, 70, 74, 75, 77 Five studies that used a reference category other than ‘none’20, 29, 43, 54, 58 were excluded from the meta-analysis. We conducted a meta-analysis of the association between presence of freckling and risk of melanoma using a ‘some’ vs. ‘none’ comparison as the data were presented in this way for the majority of studies. Of the 34 studies, 12 presented data for 2 or more categories of freckle count16, 18, 19, 21, 24, 27, 32, 55, 60, 61, 66, 74; for these studies, the categories of freckle count were combined to be included in the meta-analysis. Two studies presented data for both freckling in childhood and in adulthood66, 74; the data for adult freckling only was included in the meta-analyses.

The pooled RR for risk of melanoma associated with presence of freckling was 1.99 (95% CI: 1.79–2.20) with evidence of significant heterogeneity (Table 6). The pooled estimate was slightly lower for population-based studies than for clinic-hospital based studies (1.89 vs. 2.10). When stratified by geographic location, the pooled RR ranged from 1.42 (95% CI: 1.12–1.80) for studies conducted in Australia to 2.64 (95%CI: 1.85–3.77) for studies conducted in Northern Europe. Lower pooled RRs were observed for studies that had adjusted for other pigmentary characteristics or nevi variables.

Table 6. Meta-analysis results using a random effects model: Risk of melanoma associated with freckling
inline image

Publication bias

There was no evidence of publication bias using both the Egger weighted regression method or the Begg rank correlation method for the analyses of eye colour, blond hair, light brown hair colour or skin phototypes II or III. There was evidence of publication bias using the Egger method but not the Begg method for the analyses of skin phototypes II and III, and there was evidence of publication bias using both methods for the analysis of red/red-blond hair colour, freckling and skin phototype 1 (Supporting Information Table 1).

Population-Attributable Fraction

Estimates of the PAF associated with eye colour, hair colour, skin phototype and presence of freckling, calculated using the summary RRs derived from the meta-analysis (for all studies and also restricted to population-based studies), and the weighted average of the prevalence estimates for population-based studies are presented in Table 7. The highest PAFs were observed for skin phototypes I/II (0.27, 95% CI: 0.21–0.31), presence of freckling (0.23, 95% CI: 0.19–0.26) and blond hair colour (0.23, 95% CI: 0.20–0.26). For eye colour, the PAF for blue/blue-grey eye colour was higher than for green/grey/hazel eye colour (0.18 vs. 0.13). The PAF of melanoma associated with red hair colour was 0.10 (95% CI: 0.09–0.11) compared with 0.23 for blond and 0.15 for light brown hair colour. There was marked variation in prevalence of hair colour across studies. Using the upper and lower values of the range of prevalence for population-based studies, the PAFs ranged from 0.03 to 0.17 for red hair, 0.03 to 0.47 for blond hair and from 0.04 to 0.23 for light brown hair colour. Prevalence of the different skin phototypes, eye colour and freckling were more homogeneous across studies.

Table 7. Estimates of the population attributable fraction (PAF) of melanoma associated with eye colour, hair colour, skin phototype and freckling
inline image

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We have systematically reviewed the most recent epidemiological data reporting the relationship between melanoma and pigmentary characteristics and conducted meta-analyses of the associations between melanoma and eye colour, hair colour, skin phototype and the presence of freckling. Our findings suggest that the pigmentary characteristics responsible for the largest burden of melanoma were skin phototype I or II (PAF 0.27), freckling and blond hair colour (PAF 0.23) and blue/blue-grey eye colour (PAF 0.18). Heterogeneity was evident for most of the pooled estimates. Whilst we examined potential sources of heterogeneity including study design, sources of cases and controls, study location, whether the phenotypic characteristic was self-reported or observer recorded and adjustment for potential confounders, no single one of these factors consistently accounted for all of the heterogeneity. The dissimilarities are most likely due to inherent differences in the individual study populations. In general, there was less heterogeneity amongst the population-based studies.

While clearly important in determining susceptibility to melanoma, pigmentation characteristics are only part of the causal mechanism for this cancer. In previous work, we have shown that the presence of large numbers of melanocytic nevi confers substantially higher risks of melanoma than any of the pigmentation factors considered here, and the estimated burden of melanoma attributable to having 25 or more common nevi was high (PAF 0.42).2 The highest PAF we observed for pigmentary characteristics was skin phototypes 1/II, equivalent to that reported previously for one of more atypical nevi (PAF 0.27).2 In contrast to the high attributable fractions for these various phenotypes, only a small proportion of melanoma cases are attributable to familial risk (PAF < 0.07).3

To our knowledge, this is the first study to systematically evaluate the PAF of melanoma associated with these pigmentary characteristics using estimates of RR derived through systematic review and meta-analysis. Other strengths of this study include our comprehensive review of the literature pertaining to pigmentary characteristics and melanoma risk, and the large number of cases and controls included in our meta-analyses with concomitant power to assess the associations, and the resultant precise estimates of risk. Several limitations must also be considered. First, the studies contributing to the pooled RR estimates are prone to several biases, including selection and recall bias. Non-differential exposure misclassification would likely bias OR estimates towards the null, which would result in an underestimate of the true PAFs.78 Selection bias due to control recruitment from dermatology clinics or hospitals in the non-population-based case–control studies would result in an attenuated effect if controls recruited in this way were more likely to have certain pigmentary characteristics (e.g., fair skin, red hair and blue eyes) than those recruited randomly from a population-based source; this would also result in an underestimate of the true PAFs.78 There are also challenges associated with the use and interpretation of PAFs,79, 80 including the interpretation of multiple competing risks, and the fact that PAFs computed separately for different risk factors are not constrained to sum to 1.0. Multi-factorial attributable fractions, however, can only be estimated reliably when the data are presented and published in this way.

An earlier systematic review and meta-analysis81 included studies conducted up to 2002, however a large number of relevant studies have entered the literature since that review was published in 2005. Our meta-analyses included 24 new studies,12–18, 20, 21, 36–50 comprising 8095 new cases that were not published at the time of the earlier analysis. These studies included 3 cohort,12–14 one nested case–control, 5 population-based case–control16–18, 20, 21 and 15 clinic/hospital-based case–control studies.36–50 Our study substantially extends previous investigations by estimating the PAFs associated with pigmentary characteristics.

Technological advances in recent decades has led to much research focused on characterizing the genetic basis underlying the natural variation in the pigmentary traits of skin, hair and eye colour in humans and their association with melanoma and other skin cancers. Control of pigmentation is complex, involving a number of major genes, several modifier genes, and environmental influences such as exposure to UV radiation. Several key pigmentation genes have been characterised. The MC1R gene is associated with pigmentary risk factors, including red hair and freckles; candidate gene approaches identified MC1R variants as low penetrance risk alleles for melanoma.82, 83 A recent meta-analysis concluded that 7 SNPs in MC1R confer significantly increased risks of melanoma,84 with ORs ranging from 1.42 to 2.45. More recently variants in the OCA2 (oculo-cutaneous albinism) gene have been linked to eye colour85 and also as low-penetrance risk alleles for melanoma.86–88 Large-scale GWAS have identified several other pigmentation genes including ASIP, TYR, SLC45A2 and TRYPT1, which confer modest risks for melanoma.89–92 While high-profile genetic research has received a great deal of attention, it is important to remember that the phenotypic traits themselves, which as we have shown are responsible for a large burden of melanoma in the population, are readily observed by primary care physicians independently of information on a person's genetic risk factor profile. Two recent studies have shown that the addition of information on genotype has at best a modest ability to improve melanoma risk prediction once phenotype is known.93, 94

In terms of melanoma prevention, we have demonstrated that phenotypic assessment of melanoma risk can be informative. In particular, PAFs are helpful for estimating the burden of disease occurring within sub-groups of a population. We have identified individuals with skin phototypes I or II, freckling, blond or red hair colour, and blue/blue-grey eye colour as high risk groups who constitute a disproportionate share of patients with melanoma. We were not able to separately estimate the independent effects of these traits, some of which we know to be highly correlated, since very few studies publish risk estimates for combinations of phenotypes. Measuring the combined effects of these pigmentation characteristics with each other, and with other risk factors (especially nevi) is necessary if valid risk prediction tools are to be developed, since it is likely that interactions between factors may occur to greater or lesser degrees. Some progress has been made in this regard. For example, a recent pooled analysis of melanoma in women observed a higher risk of melanoma associated with high nevus counts among women with red hair than for women with freckling but no red hair, or women without any of these traits.95 A recent prospective study also reported a multiplicative interaction between hair colour and atypical nevi.96 Few studies have sufficient power to conduct such analyses however, and it is likely that reliable estimates of risks for combinations of risk factors will only be obtained from large prospective studies with detailed exposure measurement at baseline. This is the aim of our continuing research.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We gratefully acknowledge Patricia Williams for her assistance in preparing the manuscript. Dr. David Whiteman is a Future Fellow of the Australian Research Council.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IJC_25243_sm_supptable1.doc40KSupporting Table 1

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.