Season of diagnosis has no effect on survival from malignant melanoma

Authors

  • Harindra Jayasekara,

    Corresponding author
    1. Centre for Molecular, Environmental, Genetic and Analytic Epidemiology, Melbourne School of Population Health, The University of Melbourne, Carlton, Victoria, Australia
    2. Cancer Epidemiology Centre, The Cancer Council Victoria, Carlton, Victoria, Australia
    • Centre for Molecular, Environmental, Genetic and Analytic Epidemiology, Melbourne School of Population Health, The University of Melbourne, Level 1, 723 Swanston Street, Carlton, Victoria 3053, Australia
    Search for more papers by this author
    • Fax: +61 3 93495815.

  • Emily Karahalios,

    1. Cancer Epidemiology Centre, The Cancer Council Victoria, Carlton, Victoria, Australia
    Search for more papers by this author
  • Vicky Thursfield,

    1. Cancer Epidemiology Centre, The Cancer Council Victoria, Carlton, Victoria, Australia
    Search for more papers by this author
  • Graham G. Giles,

    1. Centre for Molecular, Environmental, Genetic and Analytic Epidemiology, Melbourne School of Population Health, The University of Melbourne, Carlton, Victoria, Australia
    2. Cancer Epidemiology Centre, The Cancer Council Victoria, Carlton, Victoria, Australia
    Search for more papers by this author
  • Dallas R. English

    1. Centre for Molecular, Environmental, Genetic and Analytic Epidemiology, Melbourne School of Population Health, The University of Melbourne, Carlton, Victoria, Australia
    2. Cancer Epidemiology Centre, The Cancer Council Victoria, Carlton, Victoria, Australia
    Search for more papers by this author

Abstract

Diagnosis in summer had been shown to be associated with better survival from some cancers, but such studies on malignant melanoma where sun exposure is a risk factor for disease are rare. We evaluated seasonality in melanoma diagnosis and its effect on survival in Victoria, Australia using 26,060 cases reported to the population-based Victorian Cancer Registry during 1986–2004. To estimate the amplitude of the seasonal variation, we calculated the ratio of the number of melanoma cases diagnosed in summer to that in winter. Linear regression was undertaken to assess the variation in thickness, the main prognostic indicator for melanoma, by season of diagnosis adjusting for sex, anatomical site, year of diagnosis and age at diagnosis. We modeled excess mortality using Poisson regression controlling for possible confounders in order to study the effect of season of diagnosis on survival. An overall 46% summer diagnostic excess was evident (summer-to-winter ratio 1.46; 95% CI 1.41, 1.52). Results of linear regression showed that melanoma diagnosed in winter were thicker than those diagnosed in any other season (percentage difference in thickness −2.01, −6.97 and −10.68 for spring, summer and autumn, respectively; p < 0.001). In the Poisson regression model of relative survival, cases diagnosed in spring, summer or autumn had slightly lower excess mortality than those diagnosed in winter before adjustment for other variables, but after adjustment the excess mortality ratios were close to unity. Our findings do not support the hypothesis that melanoma cases diagnosed in winter have worse prognosis than cases diagnosed in other seasons. © 2009 UICC

Malignant melanoma of the skin, with an estimated annual incidence of 132,000 cases globally, poses a significant public health problem in parts of the world predominantly inhabited by Caucasian populations.1 The highest incidence rate worldwide is seen in Australia, where a large proportion of the population is of European descent.2

Seasonal variation in the diagnosis of melanoma with a summer peak and a winter trough is well documented.3–11 Recent epidemiological studies indicate that people diagnosed with cancers of the breast, colon, prostate, lung and Hodgkin lymphoma during summer and autumn months have higher survival than people diagnosed during the winter.12 In a study using cancer registry data from New South Wales, Australia, Boniol et al.11 reported that melanoma cases diagnosed during the summer had higher survival than those diagnosed during the winter, an association that persisted after adjustment for melanoma thickness, which also showed seasonal variation.

In the present study, we analyzed data from the Victorian Cancer Registry to investigate the seasonal variation in the diagnosis of melanoma and its effect on survival in Victoria, which is situated in the temperate part of Australia, further away from the equator than New South Wales.

Material and methods

Study population

The Victorian Cancer Registry has been population based since 1982, registering all malignant neoplasms including invasive malignant melanoma of the skin (i.e. Clarke's level II and greater)13 notified by the hospitals and pathology laboratories in Victoria. Out of 27,473 cases diagnosed with invasive melanoma during the 19-year period 1986–2004, 1,413 with multiple primary melanomas were excluded, and the remaining 26,060 analyzed for seasonal patterns in diagnosis and survival. Data from 1982 to 1985 were not included in the current analysis due to incomplete reporting of melanoma thickness. Data on age, sex and melanoma site and thickness were extracted for individual cases.

Statistical analysis

To estimate the amplitude of the seasonal variation in diagnoses, we calculated the ratio of the number of melanoma cases diagnosed in summer to that in winter.14 Separate analyses were carried out for cases diagnosed during four consecutive time periods (1986–1990, 1991–1995, 1996–2000 and 2001–2004), to compare potential differences in the seasonal pattern over time. We pooled data for three contiguous months to accumulate the incidence for each season, considering summer to be December–February; autumn: March–May; winter: June–August and spring: September–November.

To assess the variation in thickness by season, linear regression was undertaken on the natural log of thickness adjusting for season, sex, anatomical site, year of diagnosis and age at diagnosis. The assumptions underlying linear regression were investigated. Results of the linear regression are presented as percentage difference in thickness for a unit increase in a continuous variable (age and year of diagnosis) or a percentage change in thickness comparing one category to the designated reference category for categorical variables (season, sex and anatomical site).

To study the effect of season of diagnosis on survival, relative survival was computed using the period method.15 Relative survival is defined as the observed survival in the case group divided by the expected survival of a comparable group from the general population.16 Expected survival was estimated using the Ederer II method from Victorian population life-tables stratified by age, sex and calendar time.17 Poisson regression was used to model excess mortality, which allowed us to estimate the effect of age, sex, anatomical site, thickness and year of diagnosis while controlling for possible confounding effects.18 The estimates from the model are interpreted as excess mortality ratios, i.e. an excess mortality ratio of two for males in comparison with females indicates that males experience 100% higher excess mortality than females following a diagnosis of melanoma.

All statistical analyses were performed using the software package Stata 10.0.19

Results

More than half the cases were male (Table I). The mean age at diagnosis was 56.5 years. For males, the most common site was the trunk (38.2%) whereas for females, the majority of the lesions affected the upper or lower extremity (60%) (data not shown in table). Breslow's thickness was measured for 22,547 (86.5%) of cases; the median thickness was 0.70 mm (interquartile range 0.40–1.4 mm).

Table I. Characteristics of Cases Diagnosed With Malignant Melanoma in Victoria, Australia 1986–2004
CharacteristicCases (%)
  1. Summer was considered to be December–February; Autumn, March–May; Winter, June–August; Spring, September–November.

  2. SD, standard deviation; IQR, inter quartile range.

Sex
 Male13,412 (51.5)
 Female12,648 (48.5)
Age at Diagnosis (yr)
 Mean Age (SD) 56.5 (17.7) yr 
 <405,054 (19.4)
 40–494,277 (16.4)
  50–594,589 (17.6)
 60–694,939 (19.0)
 70+7,201 (27.6)
Anatomical Site
 Head and Neck4,961 (19.1)
 Trunk7,433 (28.5)
 Upper Limb and Shoulder5,409 (20.7)
 Lower Limb and Hip6,440 (24.7)
 Overlapping Sites14 (0.1)
 Unknown1,803 (6.9)
Breslow's Thickness (mm)
 Median Thickness (IQR) 0.70 (0.4–1.4) mm 
 ≤115,183 (58.3)
 >1 and ≤23,648 (14.0)
 >2 and ≤42,318 (8.9)
 >41,398 (5.3)
 Missing3,513 (13.5)
Season of Diagnosis
 Summer7,716 (29.6)
 Autumn6,351 (24.4)
 Winter5,373 (20.6)
 Spring6,620 (25.4)
Total26,060 (100.0)

More cases were diagnosed during summer than winter (29.6% vs. 20.6%). A ‘summer-to-winter’ ratio of 1.46 [95% confidence interval (CI) 1.41, 1.52] was observed with cases with overlapping or unknown site excluded. Higher ratios were seen for females (1.54; 95% CI 1.46, 1.63), cases aged less than 40 years (1.73; 95% CI 1.60, 1.88) and 60–69 years (1.55; 95% CI 1.42, 1.68), lower (1.77; 95% CI 1.65, 1.90) and upper extremities (1.57; 95% CI 1.45, 1.69), and for lesions with a thickness of 1 mm or less (1.59; 95% CI 1.52, 1.66). This summer excess in diagnosis was evident for each time period (1986–1990, 1991–1995, 1996–2000 and 2001–2004) although the ratio declined from 1986 to 1990 (1.57; 95% CI 1.44, 1.71) and 2001–2004 (1.24; 95% CI 1.16, 1.33) (p value for trend over time < 0.001).

To further assess the variation in thickness by season, we performed a linear regression analysis on the natural log of thickness, adjusting for season, sex, anatomical site, year of diagnosis and age at diagnosis. The natural log of thickness was normally distributed and the residuals were also normally distributed. Table II shows the percentage difference in thickness adjusting for season, sex, anatomical site, year of diagnosis and age at diagnosis. Comparatively thicker lesions were seen for males, with advancing age at diagnosis and for lesions of the lower limbs. Melanomas diagnosed in winter were thicker than those diagnosed in any other season.

Table II. Linear Regression Analysis of Breslow's Thickness of Malignant Melanoma Diagnosed in Victoria, Australia 1986–2004
 Percentage difference in thickness (95% CI)p value
  • Linear regression on the natural log of thickness adjusting for season, sex, anatomical site, year of diagnosis and age; 4,034 out of 26,060 observations with overlapping and unknown sites of lesion or with missing data for Breslow's thickness were excluded from the analysis.

  • Summer was considered to be December–February; Autumn, March–May; Winter, June–August; Spring, September–November.

  • CI, confidence interval.

  • 1

    Percentage change per year in comparison to 1986, which is the reference year.

Year of Diagnosis (per yr)10.04 (0.04, 0.04)<0.001
Age at Diagnosis (per 10 yr)10.90 (10.13, 11.68)<0.001
Sex
 FemaleReference<0.001
 Male13.85 (11.01, 16.76)
Anatomical Site
 Head and NeckReference<0.001
 Trunk−9.80 (−12.92, −6.57)
 Upper Limb and Shoulder−5.93 (−9.37, −2.37)
 Lower Limb and Hip7.45 (3.56, 11.48)
Season of Diagnosis
 WinterReference<0.001
 Spring−2.01 (−5.43, 1.53)
 Summer−6.97 (−10.11, −3.71)
 Autumn−10.68 (−13.83, −7.42)

Table III shows both adjusted and unadjusted excess mortality ratios modeled using Poisson regression. When adjusted for other variables (year of diagnosis, sex, age at diagnosis, thickness, anatomical site and season) males experienced approximately a 50% higher excess mortality than females. A doubling in thickness of melanoma (e.g. from 0.5 to 1 mm or 1 mm to 2 mm) was associated with an increase in excess mortality of 2.29-fold (2.19–2.38). Before adjustment for other variables, cases diagnosed in spring, summer or autumn had slightly lower excess mortality than those diagnosed in winter, but after adjustment the excess mortality ratios were close to unity. The association between anatomical site and survival was also weakened after adjustment. A further analysis was performed excluding melanomas of the head and neck. The adjusted excess mortality ratios for each season differed by less than 10% from those of the model that included all melanomas (data not shown in table).

Table III. Poisson Regression Analysis of Excess Mortality from Malignant Melanoma in Victoria, Australia 2000–2004
 Excess mortality ratio (95% CI)
Unadjustedp valueAdjustedp value
  • Poisson regression on excess mortality; 4,034 out of 26,060 observations with overlapping and unknown sites of lesion or with missing data for Breslow's thickness were excluded from the analysis.

  • Summer was considered to be December–February; Autumn, March–May; Winter, June–August; Spring, September–November.

  • CI, confidence interval.

  • 1

    Reference year is 1986.

  • 2

    Breslow's thickness has been entered as log2(thickness). Therefore, the excess mortality ratio represents the relative risk for a doubling in thickness (e.g. comparing thicknesses of 0.5 and 1 mm or 2 and 4 mm).

Year of Diagnosis (per yr)11.01 (0.99, 1.03)0.2431.02 (1.00, 1.03)<0.05
Age at Diagnosis (per 10 yr)1.39 (1.33, 1.45)<0.0011.17 (1.13, 1.21)<0.001
Sex
 FemaleReference<0.001Reference<0.001
 Male2.40 (2.05, 2.80)1.50 (1.32, 1.70)
Breslow's Thickness (per log2[mm])22.52 (2.42, 2.62)<0.0012.29 (2.19, 2.38)<0.001
Anatomical Site
 Head and NeckReference<0.001Reference<0.001
 Trunk0.63 (0.53, 0.76)1.14 (0.98, 1.33)
 Upper Limb and Shoulder0.40 (0.31, 0.50)0.76 (0.64, 0.92)
 Lower Limb and Hip0.42 (0.34, 0.52)0.88 (0.74, 1.04)
Season of Diagnosis
 WinterReference0.200Reference0.553
 Spring0.86 (0.69, 1.06)0.92 (0.78, 1.08)
 Summer0.80 (0.65, 0.98)0.96 (0.82, 1.12)
 Autumn0.83 (0.69, 1.06)1.02 (0.87, 1.21)

Discussion

There was a pronounced seasonal variation in the diagnosis of melanoma with a summer diagnostic excess predominantly characterized by thinner lesions, mainly in females and younger patients, and localized to the extremities. We found male sex, advancing age at diagnosis, and having a lesion on the lower limbs or diagnosed during winter to be associated with tumor thickness. Season of diagnosis had no effect on survival after adjusting for other factors including thickness of the lesion.

The summer diagnostic excess is well known with two competing hypotheses used to explain this observation: enhanced ascertainment of cases because of people having greater awareness of skin lesions during summer or a late tumor-promoting effect resulting from increased ambient ultraviolet radiation.3–11

Much attention has been focused recently on the possible effect of vitamin D on cancer survival. There is some evidence that people diagnosed with cancers of the breast, colon, prostate and lung, or with Hodgkin lymphoma during the summer, when vitamin D levels are high,20, 21 have better prognosis than do patients diagnosed during the winter.12 Boniol et al.11 reported a hazard ratio for melanoma-specific fatality of 0.82 (95% CI 0.72, 0.94) comparing patients diagnosed during summer with those diagnosed during winter after adjustment for tumor thickness and other potential confounding variables. They postulated that the greater sun exposure around the time of diagnosis has a positive effect on survival either through increased vitamin D synthesis or an alternative direct anti-mutagenic effect of sunlight on skin as proposed by Berwick et al.22 who found higher survival for melanoma patients with greater lifetime sun exposure.

Our findings from Victoria do not confirm season of diagnosis as a predictor of survival. Because Victoria is at higher latitude than NSW, and has greater seasonal variation in UVB irradiance,23 if exposure close to the time of diagnosis is important, one might expect season of diagnosis to have a stronger association with survival in Victoria than in NSW. Different statistical methods were used in the two studies—we used relative survival, whereas Boniol et al. used melanoma-specific survival—but these methods should give similar results since they are both measures of net survival. Adjustment for tumor thickness, which differed substantially by season in both studies, had a greater impact in our data. Before adjustment, there was weak evidence of higher survival for those diagnosed in summer in Victoria, but this reversed direction after adjustment, whereas in NSW the association weakened only slightly and remained significant. The median tumor thickness was the same in both studies (0.70 mm), therefore making it unlikely that differences in the distributions of tumor thickness could explain the difference. We had no information on histopathological subtype, which is not recorded by the Victorian Cancer Registry, but it is unlikely that this variable would confound the relationship after adjustment for tumor thickness and the results were unchanged after excluding melanomas of the head and neck.

In conclusion, our results do not support the hypothesis that sun exposure around the time of diagnosis influences prognosis of melanoma patients. Consequently, they also provide no support for the hypothesis that vitamin D levels at time of diagnosis influence survival after diagnosis with melanoma.

Ancillary