Association between melanoma thickness, clinical skin examination and socioeconomic status: Results of a large population-based study
Survival from melanoma is inversely related to tumour thickness and is less favorable for those in lower socioeconomic (SES) strata. Reasons for this are unclear but may relate to a lower prevalence of skin screening. Our aim was to examine the association between melanoma thickness, individual-level SES and clinical skin examination (CSE) using a population-based case-control study. Cases were Queensland (Australia) residents aged 20–75 years with a histologically confirmed first primary invasive cutaneous melanoma diagnosed between January 2000 and December 2003. Telephone interviews were completed by 3,762 cases (77.7%) and 3,824 (50.4%) controls. Thickness was dichotomized to thin (≤2 mm) and thick (>2 mm). Compared with controls, the risk of thick melanoma was significantly increased among men [relative risk ratio (RRR) = 1.56, 95% CI = 1.22–2.00], older participants (RRR = 1.76, 95% CI = 1.10–2.82), those educated to primary level (RRR = 1.70, 95% CI = 1.08–2.66), not married/living as married (RRR = 1.47, 95% CI = 1.15–1.88), retired (RRR = 1.39, 95% CI = 1.01–1.94) and not having a CSE in past 3 years (RRR = 1.45, 95% CI = 1.12–1.86). There was a significant trend to increasing prevalence of CSE with higher education (p < 0.01) and the benefit of CSE in reducing the risk of thick melanoma was most pronounced among that subgroup. There were no significant associations between cases with thin melanoma and controls. Melanoma thickness at presentation is significantly associated with educational level, other measures of SES and absence of CSE. Public health education efforts should focus on identifying new avenues that specifically target those subgroups of the population who are at increased risk of being diagnosed with thick melanoma.
Melanoma is one of the most common invasive cancers in Australia with 10,326 people diagnosed and 1,238 dying from this disease in 2006 for a population of nearly 21 million. The lifetime risk (to age 85 years) of being diagnosed with melanoma is one in 18.1 In the United States, with a population of 304 million, it has been estimated that over 68,000 people will be diagnosed with melanoma and over 8,600 will die from this disease in 2009.2 The incidence of melanoma continues to increase in most countries with predominantly fair-skinned populations.3 Although mortality from melanoma appears to have stabilized over the past decade in Australia, mortality among men continues to increase in the United States.4
Tumour thickness is an important prognostic marker, with the 5-year relative survival decreasing dramatically as thickness increases.5 Having had a whole-body clinical skin examination (CSE) in the past 3 years has been shown to significantly reduce the risk of thick melanoma.6 Factors related to increased risk of thick melanoma include low socioeconomic status (SES)7 and lower educational attainment.8–11 In line with this, melanoma survival rates are generally poorer among people with lower SES. In Queensland, 5-year relative survival for those in the most disadvantaged group was 30% lower than those in the middle SES group.12 A recent U.S. study found significant differences in the mortality to incidence ratio (MIR) according to area level measures of SES. Specifically, those diagnosed with melanoma living in areas with lower levels of education had a high MIR (equivalent to lower survival), whereas a lower MIR was observed for patients living in areas with higher median incomes.11
To date, few studies have used individual-level data to measure the association between SES and melanoma thickness. The majority of studies have measured SES at a group level or have been limited to particular population groups.8–11, 13, 14 Although studies such as these are helpful, they do, however, have well-documented limitations.15 For example, they are unable to determine if any variations found are due to the characteristics of the individuals or the characteristics of the geographical areas in which they live. Currently, it is unclear to what extent individual level measures of SES are associated with melanoma thickness.
The aim of this article is to examine the association between individual level measures of SES such as education and employment status and history of CSE and melanoma thickness in Queensland, the region with highest incidence of melanoma in the world.16, 17
Material and Methods
We have previously presented data from this case-control study, for which a full description of the study methodology has been published.18 Ethical clearance for this study was obtained from the University of Queensland Ethical Review Committee, with informed consent obtained from each participant.
Briefly, eligible cases were Queensland residents aged 20 to 75 years diagnosed with histologically confirmed first primary invasive cutaneous melanoma between January 1, 2000, and December 31, 2003. In Queensland, histopathology is conducted within centralized pathology services. It is normal practice for pathologists in Queensland to seek further review for difficult lesions. For sampling efficiency and cost, all patients with thick melanoma (>0.75 mm) and a random 60% sample of those with thinner melanoma (≤0.75 mm) were included. Patients with metastatic disease, a previous melanoma, acral lentiginous or noncutaneous melanoma were excluded. Patients received an invitation to participate after permission to contact was obtained from their treating physician.
Of 4,839 eligible patients, physician consent was obtained for 4,510 (93.2%). Of these, 3,887 patients (80.3% of the total eligible sample) agreed to participate, with 3,762 (77.7% of the total eligible sample) completing an interview. The median time between diagnosis and interview was 5 months (range, 1–26 months). Females were more likely to participate than males (82.7 and 78.7% respectively, p < 0.01). Eligible cases with thicker melanoma (≥2.00 mm) were significantly less likely to participate compared with cases with melanoma <2.00 mm (72.3 and 82.3% respectively, p < 0.001). There was no difference in rates of participation according to age.
Eligible controls were selected from the Queensland Electoral Roll using stratified sampling, based on the 5-year age groups and sex distribution of the melanoma cases. Those with a confirmed diagnosis of melanoma were replaced. Of the 7,594 eligible controls, 3,972 agreed to participate and 3,824 interviews were completed (50.4%).
Data were collected from participants via a telephone interview conducted by trained interviewers using a computer-assisted telephone interview system. The interview collected information on demographics (age and sex), sociodemographics, melanoma risk factors such as skin type, skin sensitivity, moliness, previous history of nonmelanoma skin cancer and screening history (self and clinical).
Melanoma risk factors
Risk factors assessed in this study included ethnicity (United Kingdom, European and other), natural hair colour at age 21 years (blonde/red and black/brown), eye colour (blue/grey, green/hazel and black/brown), colour of skin before tanning (very fair/fair and olive/brown/Asian), tendency to burn when exposed to sun for an hour without protection (sunburn and no sunburn), degree of freckling that was based on previously sent pictorial prompts (none/few and some/many), self-reported number of moles on the back (none, 0–10, 11–30, 31–50, 50 or more), childhood sunburn experience (never, up to 10 times and more than 10 times) and age first arrived in Australia (born in Australia, 1–19 years, 20–39 years and 40–75 years). Participants were also asked whether they have first-degree relatives with a history of melanoma (yes/no) or a nonmelanoma skin cancer (yes/no).
Indicators of SES
Indicators of SES included the participant's highest level of education (primary, years 7–10, years 11, 12 or vocational and tertiary); current employment status (full-time, part-time, not working or retired) and marital status (married/living as married and separated/divorced/widowed or never married).
Clinical Skin Examination
History of whole-body CSE in the 3 years before the first signs or symptoms of the melanoma (or before a corresponding date for controls) was collected from all participants. For cases, this excluded the initial examination by a doctor that was part of the diagnostic process, unless it was a whole-body CSE of an asymptomatic patient. As reported previously,19 we undertook a test-retest reliability of the telephone interview among 164 cases and 104 controls. Reliability was good for most measures including the question relating to CSE for both cases (kappa = 0.71) and controls (kappa = 0.79) with concordance values of 87 and 90%, respectively. There was no significant difference in recall of CSE for cases according to melanoma thickness.
Interview data for melanoma cases were supplemented with pathology data held by the Queensland Cancer Registry, including tumour thickness. Thickness was collapsed into two categories (≤2 and >2.00 mm) based on the current recommended cut points within the T classification for melanoma thickness.20
For sampling efficiency and cost, a random sample of 60% of very thin melanomas (≤0.75 mm) was selected, with complete enumeration of melanomas > 0.75 mm. Therefore, we weighted the proportion of very thin melanomas in all analyses. Control participants were assigned a weight value of 1.
The relationship between melanoma thickness and the variables of interest (education, work status, marital status and prior CSE) was assessed using a multinominal regression model. The potential confounders included in the analysis were eye colour, hair colour, skin colour, degree of freckling, number of moles on the back, history of other type of cancer and history of melanoma or other skin cancer in a blood relative and ethnicity. A multinomial model (main effects) containing all these potential confounders as well as age, gender, education level, work status and marital status was fitted to the data, using melanoma thickness (two thickness categories plus the control group) as the outcome variable. Results from this model are reported in the form of relative risk ratios (RRRs). Records with missing values for the outcome or any of the explanatory variables were deleted from the analysis.
We conducted a secondary analysis to consider the association between CSE and the selected measures of SES: education level, marital status and employment status. We also included the same additional potential confounders as in the main effects analysis. Based on the results of this secondary analysis, we reran the main effects analysis, including the significant interaction terms between the SES variables and CSE to test whether CSE may be an important effect modifier in the association between melanoma thickness and SES. All data were analysed using STATA version 11 (STATAcorp, TX).
There was information available for analysis on 3,762 respondents with invasive melanoma and 3,824 controls. The majority of the melanomas diagnosed among cases were thin 2.00 mm or less (88%).
Two-thirds of melanomas >2.00 mm occurred in men and just over half (53.6%) occurred in those aged 60 years or older (Table 1). The distribution of educational level was similar for controls and for cases with melanoma ≤2.00 mm. However, a significantly higher proportion of cases with melanoma >2.00 mm had been educated to primary level only (p < 0.001). Cases with melanoma >2.00 mm were also less likely than controls or those with thinner melanoma to be married/living as married (p = 0.014), to be working full-time (p < 0.001) or to have had CSE in the past 3 years (p < 0.001).
Table 1. Bivariate associations between melanoma thickness and sociodemographics for 7,586 study participants1
In bivariate analysis, CSE in the past 3 years was twice as prevalent for participants with tertiary education compared with those with primary education only (42.1 vs. 21.7%; Table 2). Men (32.6%) were more likely to have had a CSE than women (29.3%), and married/living as married participants (32.7%) were more likely to have had CSE than those not married (25.8%). Age group (CSE was more prevalent in those older than 40 years) and work status (those in full-time work had a higher prevalence of CSE than those not working or working part-time) also had significant bivariate associations (p < 0.05) with CSE (Table 2).
Table 2. Bivariate associations between having a clinical skin examination1 and sociodemographics for 7,586 study participants2
We examined the percentage of thick (>2.00 mm) melanoma among cases categorized according to their education level and history of CSE (Table 3). Overall, the percentage of thick melanoma decreased with increasing educational level (23% of cases with a primary level education had a thicker melanoma compared with 8.4% of cases with tertiary education). Also, overall, cases with a history of CSE had a lower percentage of thick melanoma than cases without CSE. However, the association with CSE was not uniform across educational levels. The association appeared strongest for tertiary-educated cases for whom the percentage of thick melanoma was 4.4% for those with a history of CSE compared with 11.7% for those without CSE (p < 0.05). However, there was no difference in the percentage of thick melanoma according to history of CSE for cases with a primary level education (23.7 vs. 22.7%, p = 0.236). We additionally examined the interaction between CSE and educational level separately for men and women with thick melanomas. The effect was strongest for men, in that the percentage of thick melanoma for men increased with decreasing levels of education (9.1% for tertiary to 28.2% for primary) compared with women with 13.6% for tertiary and 12.1% for primary level education (data not shown).
Table 3. Percentage of thick (>2 mm) melanoma among cases categorized according to education and history of clinical skin examination
The results from the multinomial regression model are shown in Table 4. After adjustment, male gender, older age, lower educational attainment, being retired, not married/living as married and not having had a CSE in the 3 years before diagnosis were all independently and significantly associated with an increased risk of thick melanoma, and of these, being aged 60 to 69 years (RRR = 1.76, 95% CI = 1.10–2.82) or 70 to 75 years (RRR = 1.90, 96% CI = 1.12–3.22) and lower educational attainment (RRR = 1.70, 95% CI = 1.08–2.66) were the strongest factors related to diagnosis of thick melanoma. We reran the main effects model including the interaction between education and CSE, adjusting for the same potential confounders; however, although suggestive, the interaction did not reach statistical significance (χ2 = 15.50, df = 9, p = 0.08).
Table 4. Multinomial model of associations between melanoma thickness and socioeconomic characteristics1,2
The results of this study add to the limited body of evidence showing a significant relationship between melanoma thickness at diagnosis and indicators of SES, specifically, educational level and employment status. Lower educational attainment, male gender and older age were strong independent predictors of thick lesions in this large population-based study. In addition, we found a higher risk of thick melanoma among those who were retired and among those who were not currently married or living as married. We have previously reported6 that CSE in the past 3 years significantly reduces the risk of thick melanoma, and the results here suggest that it may be a mediator between education and melanoma thickness.
In relation to measures of SES, we found that respondents with higher levels of education were less likely to present with thicker melanomas compared with those with low levels of education. Although this study was not able to ascertain the reason for this association, it is consistent with the association that education levels have generally with morbidity and mortality risks and late presentation of cancers. Education has been shown to influence an individual's position in society and stress levels,21 access to health care22 and health information,23 thereby affecting decision making and preferences. More highly educated people are less likely to smoke, exercise more, wear seatbelts more often and, importantly, are more likely to participate in screening programs for breast and cervical cancers.24 This suggests that the effect of education on health seems to be driven primarily by differences in health behaviours.25 In screening programs such as mammography, fecal occult blood testing and cervical screening, low levels of education are consistent predictors of nonattendance or nonadherence, and this is similar to what we have found with education and CSE.26–28 In our study, overall and in the absence of CSE, higher educational attainment was associated with a lower risk of thick melanomas, and it was also associated with having had a CSE in the previous 3 years.
Education can increase participation in preventive health practices such as CSE by raising awareness of the importance of undertaking regular health checkups.29 It can also improve understanding of information regarding periodic tests, communication with physicians and interpretation of any results.30 Hence, the reduced impact of CSE on melanoma thickness we found for less-educated participants may be due to their relative inability to process the specific health messages provided by the doctor during the CSE that relate to early detection and ongoing skin examination. The evidence of an association between CSE and melanoma thickness among less educated participants may be partly due to a lower prevalence in this group of other behaviours relating to early melanoma diagnosis, such as performing self-examination, having regular checkups and taking prompt action if a skin lesion develops or an existing one changes.
In this study, participants who were retired and those who were not married/living as married were at higher risk of being diagnosed with thick melanoma compared with controls. It has been reported previously that those who are not married or living in a relationship are more likely to be diagnosed with later stage cancer. Our results are similar to the few other studies that included marital status as a predictor of melanoma stage at diagnosis. Van Durme et al.31 found an approximate 53% increased risk of being diagnosed with late stage melanoma for those who were unmarried. A further study of those aged 65 years or older found that the odds of being diagnosed with late stage melanoma increased if one was single or widowed.32 Although, for our study, the relatively small number of cases who were widowed (n = 189) precluded having this group as a separate category in our multivariate analysis, we did find a trend toward a higher proportion of thick melanoma for those who were widowed on stratified analysis. However, we also found that being retired was significantly associated with increased risk of being diagnosed with thick melanoma, approximately 90% of those aged 60+ years were retired indicating that even after adjusting for age, some residual confounding may still be present in the model.
Public health education campaigns, principally Australia's long-running SunSmart program, provide extensive messages around the issues of prevention and awareness, and particularly for older individuals, early detection of skin cancer. Early detection of skin cancer is one of the key cancer prevention policies of the Cancer Council Australia, Australia's premier nongovernment cancer control organization, and is also promoted by primary care organizations such as the Royal Australian College of General Practitioners.
Some limitations do apply to this study. First, although the response rate for cases was quite reasonable (78%), our control response rate was somewhat less (51%). Although differing categories of education make it difficult to examine whether our control participants were more highly educated than the population from which they were sampled, we were able to assess that a slightly higher proportion of control participants had a tertiary education (18%) compared with the population from which they were sampled (15%).33 Moreover, our study was limited in that it did not collect information on a wider range of SES variables. However, an advantage we did have was that our measures of SES were on an individual level and not aggregated. The addition of individual SES measures such as household income, occupation categories or housing arrangements might be helpful for further assessing the impact of individual level socioeconomic measures on melanoma thickness at diagnosis. Other studies have suggested that inequalities in education might be as, or more important, than differences in occupation category.29 Finally, although it may be possible that there were some differences in recall of CSE according to time from diagnosis to interview for cases and from referent date to interview for controls, when we examined prevalence of CSE by time period, we could find no systematic differences either for cases or controls.
Our results provide evidence that, independent of CSE history, higher education is associated with a lower risk of thick melanoma at diagnosis. Because CSE is independently related to lower risk of thick melanoma,6 this association is enhanced further, as those with higher education are also more likely to participate in CSE. As survival from melanoma is strongly related to melanoma thickness, these results suggest that a thorough understanding of how education and other measures of SES impact on melanoma thickness and CSEs should be a research priority.
In our study, approximately 60% of those who indicated they had a CSE in the 3 years before their diagnosis (or referent date for controls) had this conducted by their primary care physician (similar for cases and controls). In Australia, where a universal health system exists, it is estimated that around 85% of the population sees a primary care physician at least once per year.34 Within primary health care in Australia, current guidelines recommend physicians develop surveillance programs for individuals at high risk; assess patients who are concerned and develop appropriate management programs depending on their level of risk; and identify risk factors for skin cancer in patients presenting for other reasons.35 It is additionally recommended that physicians use strategies to accommodate disadvantaged individuals, including those with lower education, which will help improve their screening behavior.
Melanoma thickness at presentation is significantly associated with educational level and other measures of SES. The higher prevalence of CSE among participants with higher education levels and the greater impact of CSEs on melanoma thickness among this subgroup suggest that current early detection and skin examination messages may be reaching this subgroup more effectively. Public health education efforts need to focus on identifying new avenues that specifically target those subgroups of the population who are at increased risk of being diagnosed with thick melanoma.